i

PRAISE FOR ARTIFICIAL INTELLIGENCE FOR HR

“I found Ben’s book to be a very useful distillation of a complex and

increasingly important area of HR capability and investment. My
students, future HR leaders, gained tremendous insight into AI for
HR and the facilitating role it will soon play in the best organizations
for people. Artificial Intelligence for HR demystifies a complex topic
and contextualizes what some regard as just the latest HR fad. It is a
well-organized and well-written book worth reading by HR leaders,
educators and students. Despite the title, it’s really all about people!”
John Haggerty, Senior Lecturer, HR Studies at Cornell University

“The HR landscape has changed dramatically over the last three

years. As the ‘War for Talent’ continues to escalate, the value of the
business-minded, technologically aware HR professional has
increased. What’s needed are HR professionals who view AI and
other technology as an augmentation to HR, giving them the ability
to be business consultants versus the more common, reactive HR.
Ben’s book explores how HR professionals are using new technology
to transform their businesses and the industry as a whole. If you are
in search of a data-packed, transformational book, I’d encourage you
to take the time to dive deep, take notes, and transform your own
business with what you learn.”
Trent Cotton, VP Talent Acquisition and Retention at Bureau Veritas and Author
of Sprint Recruiting

“Artificial Intelligence for HR is a must-read for those involved with

human resources. Ben’s dedication and passion for HR promotes
instrumental success for organizations and leaders alike. This book
offers insights into the ever-changing HR environment and naviga-
tion through new AI technology while still promoting the evolutionary
development of human resource professionals.”
Jamie McCall, Director, Talent Acquisition at The Henry M. Jackson Foundation
for the Advancement of Military Medicine
ii

“Ben does an excellent job helping HR leaders who have historically

been trained with deep departmental expertise apply new AI tech-
nologies to their domains. More importantly, he helps them transition
to strategically addressing the workforce with consumer experiences
which earn their loyalty, net promotion, retention, and engagement
the same as we experience in our personal lives.”
Randy Womack, CEO, Socrates.ai

“Today’s HR leaders must understand technology to thrive in the

modern business environment and AI is one of the most transforma-
tional technologies of our age. This is often an intimidating reality for
HR professionals since many signed up not realizing digital acumen
was a top skill for career success. Fortunately, Ben has translated his
wealth of experience and expertise into an easy-to-follow handbook.
He simplifies the complexity of AI and highlights pragmatic opportu-
nities for every HR function. It’s a must-read for any HR professional
looking to thrive in today’s disruptive environment.”
Christopher Lind, Chief Learning Officer at ChenMed and Founder, Learning
Sharks
 iii

Artificial Intelligence for HR

Use AI to support and develop
a successful workforce

SECOND EDITION

Ben Eubanks
iv

Publisher’s note
Every possible effort has been made to ensure that the information contained in this book is
accurate at the time of going to press, and the publishers and authors cannot accept responsibility
for any errors or omissions, however caused. No responsibility for loss or damage occasioned to
any person acting, or refraining from action, as a result of the material in this publication can be
accepted by the editor, the publisher or the author.

First published in Great Britain and the United States in 2019 by Kogan Page Limited
Second edition 2022

Apart from any fair dealing for the purposes of research or private study, or criticism or review, as permitted
under the Copyright, Designs and Patents Act 1988, this publication may only be reproduced, stored or
transmitted, in any form or by any means, with the prior permission in writing of the publishers, or in the
case of reprographic reproduction in accordance with the terms and licences issued by the CLA. Enquiries
concerning reproduction outside these terms should be sent to the publishers at the undermentioned
addresses:

2nd Floor, 45 Gee Street 8 W 38th Street, Suite 902 4737/23 Ansari Road
London New York, NY 10018 Daryaganj
EC1V 3RS USA New Delhi 110002
United Kingdom India
www.koganpage.com

Kogan Page books are printed on paper from sustainable forests.

© Ben Eubanks 2019, 2022

The right of Ben Eubanks to be identified as the author of this work has been asserted by him in accordance
with the Copyright, Designs and Patents Act 1988.

ISBNs
Hardback 978 1 3986 0402 5
Paperback 978 1 3986 0400 1
Ebook 978 1 3986 0401 8

British Library Cataloguing-in-Publication Data

A CIP record for this book is available from the British Library.

Library of Congress Cataloging-in-Publication Data

Names: Eubanks, Ben, author.
Title: Artificial intelligence for HR: use AI to support and develop a
successful workforce / Ben Eubanks.
Description: Second edition. | London; New York, NY: Kogan Page, 2022. |
Includes bibliographical references and index. | Summary: “Artificial intelligence is changing the world of
work. How can HR professionals understand the variety of opportunities AI has created for the HR
function and how best to implement these in their organization? This book provides the answers. From
using natural language processing to ensure job adverts are free from bias and gendered language to
implementing chatbots to enhance the employee experience, artificial intelligence can add value
throughout the work of HR professionals. Artificial Intelligence for HR demonstrates how to leverage this
potential and use AI to improve efficiency and develop a talented and productive workforce.
Provided by publisher.
Identifiers: LCCN 2021051397 (print) | LCCN 2021051398 (ebook) | ISBN
9781398604001 (paperback) | ISBN 9781398604025 (hardback) | ISBN 9781398604018 (ebook)
Subjects: LCSH: Personnel management. | Labor supply. | Artificial
intelligence.
Classification: LCC HF5549 .E828 2022 (print) | LCC HF5549 (ebook) | DDC
658.300285/63–dc23/eng/20211020
LC record available at https://lccn.loc.gov/2021051397
LC ebook record available at https://lccn.loc.gov/2021051398

Typeset by Integra Software Services, Pondicherry

Print production managed by Jellyfish
Printed and bound by CPI Group (UK) Ltd, Croydon CR0 4YY
 v

CONTENTS

List of figures and tables viii

About the author ix
Foreword by Trish McFarlane x
Preface xiii
Acknowledgements xvi

01 A snapshot of HR today 1
The age of disruption 3
Researching and understanding HR today 7
The power of the employee experience 13
The new normal: Doing more with less 16
Using this book 19
Key points 20
Notes 20

02 The basics of artificial intelligence 23

A layperson’s guide to predictive HR analytics 26
Artificial intelligence components 29
AI on a global scale 43
Key points 46
Notes 47

03 General AI applications within HCM 49

AI solves administrative HR problems 51
What about the impact on jobs? 55
Types of broad-spectrum AI components in HCM 61
Key points 78
Notes 78
vi Contents

04 Core HR and workforce management 81

Core HR is core to the business 82
Workforce management 84
Payroll 86
Benefits administration 90
Employee self-service 94
Diversity and harassment 96
Key points 99
Notes 100

05 Talent acquisition 103

Sourcing tools and technologies 109
Candidate matching 115
Assessments 118
The role of data in recruiting 120
Recruiting applications for natural language processing 121
Chatbots as a communication mechanism 124
Chatbot metrics to consider 129
Key points 130
Notes 130

06 Learning and development 133

Closing the skills gap 138
Formal learning applications for AI 140
Learning content and curation 145
Informal learning applications for AI 148
Coaching and mentoring 151
Performance support 152
Looking forward: Virtual, augmented and mixed reality
  technologies 154
Key points 155
Notes 156
 Contents vii

07 Talent management 159

Replacing humans with algorithms: Hard lessons learned 160
The critical role of managers in talent management 163
Employee engagement 165
Performance management and enablement 169
The gig economy and team development 174
Internal talent mobility and career pathing 176
Key points 182
Notes 182

08 Challenges of adopting AI technology 185

Biased systems 186
Technological complacency 191
Algorithm aversion 193
Data privacy and AI 198
System integration is lacking 200
Despite the challenges… 203
Key points 206
Notes 206

09 HR skills of the future 209

The spectrum of augmenting humans with AI 211
The importance of soft skills 216
A final question 239
Key points 241
Notes 241

Afterword 245
Index 249
viii

LIST OF FIGURES AND TABLES

Figure 1.1 Lighthouse Research & Advisory talent mobility

profit chain 9
Figure 1.2 Approach to learning content strategy 11
Figure 3.1 AI-focused HR technology firms receiving 2021
venture capital funding 54
Figure 6.1 Creating engaging learning experiences 142
Figure 7.1 Performance practices of high-performing
companies 169
Figure 9.1 Human vs. AI task decision tool 213
Figure 9.2 Breakdown of average job tasks 219
Figure 9.3 Skills of the future 221
Figure 9.4 Future soft skill importance by role 222

Table 2.1 Types of talent analytics 27

Table 2.2 A sampling of artificial intelligence technologies 42
Table 3.1 Task prioritization matrix 62
Table 5.1 A sampling of female HR technology founders with
AI-based recruiting technology 107
Table 6.1 Skill development methods, ranked 138
 ix

ABOUT THE AUTHOR

Ben Eubanks is an author, speaker and researcher. He spends his days

as the owner and Principal Analyst at Lighthouse Research & Advisory.
He has authored hundreds of reports, ebooks and other resources
over the course of his career. He also maintains upstartHR, a blog
that has touched the lives of more than one million business leaders
since its inception. He is the host of We’re Only Human, a podcast
that examines the intersection of talent and technology in the work-
place through interviews with business leaders, researchers and other
insightful guests. We’re Only Human is one of the featured shows on
HR Happy Hour, the world’s longest-running HR podcast network.
Previously an HR executive for an Inc. 500-ranked global technol-
ogy startup, he currently operates Lighthouse Research & Advisory,
a human capital research and advisory services firm. He works with
HR, talent and learning leaders across the globe to solve their most
pressing business challenges with a research-based perspective
tempered by practical, hands-on experience. He also supports tech-
nology vendors with a wide variety of content, thought leadership
and advisory services.
Most importantly, he has four children, a wonderful wife of
10-plus years and a preference for running in a variety of ­outlandish
races for fun.
x

FOREWORD

Artificial intelligence. The very words conjure up images of robots

taking over the world. And, after all the science-fiction movies over
the last forty to fifty years or so, of course they would. We’ve long
been fascinated with the idea that a robot can be like a human; that
a robot can perform human tasks. But, when you look back to the
early years of robotics, the thought of a ‘thing’ having any type of
artificial, human-like intelligence seemed like a dream for far into the
future.
Today, those dreams are now a reality. Although still in its early
stages, artificial intelligence as it relates to employees and the work
we do is quickly becoming something that human resources leaders
need to be aware of and educated about. This adds to the ongoing
arsenal of tools and topics that HR is responsible for. Do you remem-
ber ten years ago when this thing called ‘social media’ was new and
something we all had to learn from a business perspective? It seemed
very scary because of all the unknown factors. We quickly got up to
speed, researched and wrote policies and procedures to prepare.
Transparently, we also failed on some, rewrote them and re-­examined
our approaches. What we learned from that experience is HR profes-
sionals need to be keenly aware of how technological advances and
communication impact the workplace. We also have to be at least a
step or two ahead of the rest of the employees so that we provide the
insight and advice our organizations need.
Artificial intelligence in the workplace is similar. We cannot turn a
blind eye, or say, ‘We will learn about this later.’ Later is too late. The
fact that you are reading this book means that you have taken the
first of many positive steps in expanding your knowledge and opin-
ions of AI and how to use it, and it also means that you are forward
thinking in how it applies to your organization. What are the areas in
which your employees are currently not working in an effective
 Foreword xi

manner? How can technology be used to supplement the existing

workflows in order to increase productivity or engagement?
As you dive into each chapter, there are a few things to keep in
mind to help ensure you make the most of the journey. First, be open-
minded. Since the capabilities of machine learning are constantly
developing, so are our reactions to and applications of that learning.
Artificial intelligence and machine learning can be scary for many
people. Coming back to the notion of robots taking over our jobs,
employees may be under the impression that AI will help replace
them. We know that HR is cast as the watchdog of our organizations
in terms of keeping the status quo, maintaining continuity and mini-
mizing risk. This is one time where that will not benefit your
organization, your employees or your own development. Read each
chapter with a curious, open mind. Really consider whether each
suggestion or example could make you or your organization better,
more productive or more engaged.
Next, think about the impact to process and workflow.
Organizations tend to get stuck in their old ways and do not make
time to re-examine them very often. Use this opportunity to think
about the current processes you have in place and how artificial intel-
ligence can positively impact changes and improvements in the
workflows. The opportunity for you as an HR professional to posi-
tively impact these workflows is great. It opens up you and your team
to circumstances that require strategic planning. In fact, you may find
that before you adopt some of the technologies enhanced by artificial
intelligence, you need to go through a needs assessment and make
changes to existing processes and workflows. This is a common chal-
lenge with any technology selection and implementation. After all,
who wants to automate the wrong processes?
Finally, think about change management. Artificial intelligence and
how it will be applied can cause angst or concern for many people
today. As you gain understanding, think about not only how AI will
affect your organization from an operational standpoint, but how
you’ll begin to craft messages to explain the potential impacts. How
will you get buy-in from all levels of the organization? What will
each stakeholder group appreciate and be concerned about regarding
xii Foreword

a­rtificial intelligence? How can you and your team proactively

address these concerns?
These are just a few ideas to get you started down the path of
greater understanding with artificial intelligence. Now, what are you
waiting for?

Trish McFarlane
CEO and Principal Analyst, H3HR Advisors
 xiii

PREFACE

My career has led me places I never imagined going, but I wouldn’t

have had it any other way. When I was a child I wanted to work in
human resources before I even knew what it was. Seriously. For a
profession that so many people ‘fall into’, I chose HR as a career path
on purpose. For starters, I was the older middle child. My three
brothers depended on me to be the peacemaker and mediator for
disputes, and because I didn’t have any position of authority, I had to
keenly hone my influence and negotiation skills. More practically, my
parents have owned a small business since I was a child, and I always
saw the troubles they had with hiring, motivating and retaining their
people. As an entrepreneurially minded child I thought to myself, ‘If I
can just get a degree in business and find out how to solve this prob-
lem, I will forever be in demand.’ Fast forward to college, and a
professor asked me to write a paper on this thing I had never heard
of called ‘human resources’. Honestly it sounded incredibly boring,
but once I started researching what it was, I realized that HR was
exactly what I had always wanted to do with my life and my career.
I graduated with my degree in human resources management and got
to work.
I started my career, as a lot of HR professionals do, dealing with
stacks and stacks of employee files and all of the administrative fun
that comes with that. Over time I worked my way through several
positions until I was working as an HR executive at a fast-growing
technology startup, and the challenges and pace of business kept me
on my toes. I gathered my HR certifications along the way and
maintain them to this day, a tribute to my roots and a reminder that
in spite of all the neat things technology can do, it will never replace
the capability of great people. Throughout this period of time, I had
been writing and blogging as a side hobby about my experiences
working and learning in HR, and a friend reached out to see if I
xiv Preface

wanted to take on a role as an analyst in the HR industry. The way I

remember her explaining it was, ‘You already like writing and speak-
ing, but now this can be your job. Plus, you get to do research.’ I was
sold. I moved into an analyst firm and soaked up everything I could
about the industry before moving on to run my own firm, which is
where I am today.
I have the most amazing job, if I’m going to be honest. Each day
presents opportunities to connect with leaders from across HR and
business, helping them achieve their goals. Additionally, I spend a
considerable amount of time learning from the vendors about their
technologies and how they operate, often advising them on product
strategy or market trends. The research I get to do every day excites
me, and it gets me closer to my life’s mission: making HR better, one
HR pro at a time. Every time I educate someone, including yourself,
it is the proverbial rising tide that lifts all boats. We all get better
when one of us gets better.
However, this book isn’t about my past or my biography. That’s
short enough to fit into a few paragraphs above, for the most part.
This book is about one of the research efforts I’ve undertaken that I
believe has the chance to fundamentally change the HR profession in
ways that have never before been seen. Technology has always been
a part of how HR operates. While software is the focus of this book
and is also what we think of when we discuss technology, it’s impor-
tant to remember that even paper is a technological improvement
over raw memory. I believe we are on the cusp of an incredible shift
in the technology landscape that has the capability to affect how HR
gets done at companies around the world. Artificial intelligence tech-
nologies are shaking up industries like healthcare, cybersecurity and
retail. However, those changes are also affecting specific functions in
the business, such as marketing, customer services and business oper-
ations. There are tremendous amounts of resources out there focusing
on some of those other aspects of how artificial intelligence will
change industries or functions, but when I started this research
project there was nothing clear cut about how this will affect the
profession of HR that I know and love.
 Preface xv

The field of HR really needs a special touch. It’s not just enough to
understand the technology. Lots of people are technical in nature and
can’t relate it back to the human impact on the workplace. The right
approach is to look at the problem from the other end of the spec-
trum: what challenges do businesses face when dealing with candidates,
employees and non-employee talent, and how can technology help to
resolve those? I’d like to think that my background as an HR execu-
tive helps to ground me and my research so that it takes on a very
practical flavour. I’m not interested in hype. I’ve had to go to my CEO
with people-related challenges and know that there is no margin for
putting your faith in an unreliable or unrealistic technology. I’m also
not an academic – any research effort I take on needs to have the
potential to improve the lives of HR, talent and learning leaders by
giving them better information or tools to improve their performance
and results. Perhaps most importantly, my fundamental belief is that
artificial intelligence has the opportunity to make HR more human,
not less. I’ll explore that notion at some level in virtually every chap-
ter, saving the deep dive for the final chapter, where I talk about the
skills you need for success in the era of artificial intelligence.
This book has been written with these concepts in mind. I’ve gone as
far as possible to make it practical, insightful and helpful. I’ve provided
examples and case studies everywhere I could to give you the insights
you need to apply the concepts and principles the chapters explore, not
just consume them and move on. And, as I’ve realized, the technological
landscape is always changing. Some of the companies and functionali-
ties I talk about in this book will change over time, and that’s fine. It’s a
fact of life. That’s the warning label you need when writing a book
about a cutting-edge technology that is still evolving rapidly, but I did
my best to clarify what technologies are available today and what tech-
nologies I envision becoming available in the future.
I know your time is valuable, so I want to close by thanking you
for your time and attention in reading this book. Based on the feed-
back I’ve received from those that have heard me talk about the
concepts within the following pages, it will be well worth your time.
I’m writing about this topic because I’m passionate about it and
passionate about improving the field of HR.
xvi

ACKNOWLEDGEMENTS

This book has been quite the undertaking and I have done my very
best to ensure that I give appropriate credit to everyone within the
context of each chapter who offered insight or ideas that supported
the content. That said, there’s always someone who is left out of the
final product, so I’ll just say this: if you and I have interacted,
connected or collaborated over the course of my career, there’s a good
chance some measure of that influence made its way into this book. I
am who I am because of the people I surround myself with, and I’d
like to think that I surround myself with good kinds of people. To
each of you, I say thanks!
Two of the special people I can’t imagine being here without:

Trish, you’ve taught me more about collaboration than I could have

ever learned on my own. I appreciate you.

Christine, your leadership and guidance helped me early in my career

and prepared me for the path I’ve taken. You’re still the best boss I
ever had.
 1

01

A snapshot of HR today

In 2014, Carnegie Mellon had some of the world’s brightest robotics

minds working on its campus. These scientists were focused on the
bleeding edge of robotics, improving upon the capabilities of existing
technology and leveraging research to design new systems. Their
insights, developed within a university setting, could have created
new breakthroughs and advancements in the use of robotic technol-
ogy for the betterment of mankind.
But then, virtually overnight, they all left.
In a surprise move, Uber lured the scientists away, hiring a signif-
icant portion of Carnegie Mellon’s robotics talent in a single
sweeping move. This hiring spree was a step forward in the firm’s
strategic pursuit of a self-driving fleet of automobiles, but it clearly
caught Carnegie Mellon by surprise. Looking back, it’s hard to
blame the educational institution for not seeing this upstart tech-
nology firm as a threat. Logically, when Carnegie Mellon considered
its peers in the marketplace, it evaluated other top-shelf universities
and research think tanks as competitive threats. However, it didn’t
consider the possibility of a ride-sharing service snagging its highly
prized robotics talent.
What this means for today’s business leaders is that it’s no longer
enough to simply look at the two or three long-term competitors to
keep a pulse on the industry. New competition can come from any
direction. And company leaders are funny in that they always think
that they are somehow shielded from the impact of these new tech-
nologies and business models. It’s a great example of what happens
2 ARTIFICIAL INTELLIGENCE FOR HR

when firms only consider their traditional competitors and markets

instead of broadening their perspective. This is a small example of
disruption and how it impacts the workplace, but it signifies a
larger change that leaders must take seriously in order to remain
competitive.
This story offers a clear understanding of the concept of industry
convergence, just one of the many trends shaking up the workplace
today. In a recent global study of chief human resources officers
(CHROs), IBM found that the number one concern for these talent
leaders was industry convergence.1 A good example of this comes
from the competition between Carnegie Mellon and Uber,2 but others
exist as well. For example, in a discussion with a Canadian banking
and financial services employer with tens of thousands of employees,
one of the HR executives was quick to redirect the conversation
when industry was mentioned. He claimed the firm was a technology
company first and a financial services institution second. This conver-
gence means that more employers are looking to hire technical
engineering talent, even if they aren’t in traditional technology indus-
tries, tightening the labour market for sought-after software engineers.
Look around you. Today’s human resources practice is not the
same as that of years past. More and more companies are looking for
ways to leverage their people as a strategic differentiator, giving them
an edge over the competition. New technologies and tools are shift-
ing the conversation for HR and business leaders, enabling them to
have greater insights into organizational functions, outcomes and
variables. Yet, in spite of all of this, there are challenges that we
simply can’t seem to shake. Globalization is forcing employers to
change how they have traditionally approached markets and talent.
Disruption is affecting businesses in a wide spectrum of ways, from
changes in consumer preferences to radical departures from tradi-
tional business models and methods. Virtual and digital work is
replacing some of the physical environments for some jobs, creating
different challenges for engaging employees than required in the past.
Plus, with a renewed focus on racial justice and equity, workers are
demanding that employers level the playing field to ensure a fair and
equitable experience for all. And, with greater demands for delivering
 A SNAPSHOT OF HR TODAY 3

memorable, engaging employee experiences, human resources (HR)

leaders are stressing about doing more with less. New technologies
like artificial intelligence (AI) seem to offer some glimpse of hope, but
many HR and business executives find themselves asking how these
tools work and what they actually do. Are they marketing hype? Are
they actual solutions to everyday problems? What can help busi-
nesses survive and thrive in a competitive market full of disruption?
Within the scope of this book, we will address each of those ques-
tions, clarifying what is and isn’t possible with the technologies
available today.

The age of disruption

The age of disruption is a blend of multiple competing priorities. Not
only does this idea of industry convergence factor into the puzzle, but
the way employers acquire talent is shifting as well. Employers are
moving away from the traditional ‘buy and hold’ approach to a more
flexible, nimble version based on project work and a variety of non-
traditional worker types. Yes, for as long as there have been workers
there have been side gigs, which can include freelancing, contracting
or other short-term assignments. However, while the concept of ‘gigs’
isn’t anything new, the platforms available today give workers more
and more control over how their time can be exchanged for money. In
economics terms, it reduces the friction of the labour market, making
it easier to match up skills with those that want to pay for them. Fiverr
allows designers and other workers to offer their skills in small projects
with a range of compensation depending on their experience and
quality. Task Rabbit allows users to hire someone for small tasks, such
as a quick grocery store trip or coming in to an office to file some
paperwork. And the popular ridesharing apps now give virtually any
consumer the ability to push a button and have a vehicle appear
within minutes to take them anywhere they want to go. From an
employer perspective, applications like Wonolo and GigSmart do the
same thing – they essentially enable employers to push a button and
have contract workers show up ready to work. While their workload
4 ARTIFICIAL INTELLIGENCE FOR HR

is often around more labour-intensive or low-level tasks, it’s a fore-

shadowing of the future when more and more work can be handled
by non-employee talent. Plus, this whole gig segment of the workforce
is increasing rapidly. Research from Princeton University shows that
the platform gig economy grew tenfold between 2012 and 2016.3
While this seems like a small piece of the overall economy, if these
platform gig workers were aggregated together as a single employee
population of 800,000 workers, the company would be larger than
Target and General Electric combined.4
This doesn’t even take into account the changes in data and tech-
nology. Today, firms collect data on virtually everything. Every
interaction. Every connection. All of this information, aptly termed
‘big data’, is being gathered, yet employers still struggle to make sense
of the information in ways that can add value to the organization.
This is true even more so for employee-focused data, because while
marketing and sales tools have been more insight-driven over time,
HR and talent systems are just reaching this level of maturity. In the
Lighthouse Research & Advisory 2020 Business Agility and HR
Technology Study, we found a variety of interesting research insights.
For instance, business leaders working outside HR are more likely
than HR leaders to see human resources technology as a strategic
tool, not just an administrative one.5 That study also pointed out a
powerful figure: 97 per cent of agile firms say that their HR technol-
ogy plays a role in enabling that future readiness. Data takes a central
position in the discussion in the following chapters due to its role in
empowering the technologies that businesses leverage today.
When we layer in increasing instances of virtual and remote work,
we also see greater pressure on employers to create, deliver and meas-
ure employee experiences and impacts at scale. Whether an employer
has 10 employees or 10,000, more and more companies are thinking
about or have already moved into a more virtual arrangement.
During the Covid-19 pandemic, several high-profile firms committed
to allowing their teams to work remotely on an indefinite basis. While
the practices of these types of firms are outliers in the employment
community, their approaches often serve as a signal to the broader
 A SNAPSHOT OF HR TODAY 5

market about what is next for the majority of employers. Virtual

work introduces the need to manage and support employees that
may never actually see each other physically in person. While that
seems simple from a technology standpoint, it also has emotional and
psychological components. For instance, do you trust someone as
much when you’ve only met them virtually, or do you trust someone
more that you’ve actually met in person? When that is multiplied
across the entire workforce, it helps to paint a picture of the impor-
tance of being intentional about remote and virtual work.
On the worker side of the equation, employees also now have
greater expectations of the technology they access and utilize in the
workplace. Consumer technology preferences for intuitive software
applications that are mobile-friendly and available anytime, anywhere
are now commonplace. This is intriguing because even as recently as
five or ten years ago, the bulk of the workforce didn’t have any real
say-so in how companies adopted and leveraged HR technology. In
fact, virtually all of the technology adopted was done so purely with
the administrators in mind. The question asked when looking at tech-
nology solutions was, ‘What makes it easier to do HR, learning or
recruiting work?’ However, today employees often have access to
self-service tools for a range of needs, from tracking competencies
and performance to updating personal information. That means
employees now have yet another lens through which to evaluate their
employers. As we’ll see in the coming chapters, some of those tools
are powered by artificial intelligence technologies that can offload
HR tasks and improve user satisfaction at the same time. And if it
sounds like a relatively minor item, our research shows otherwise. This
type of technology can actually influence the employee experience –
our research shows that high-performing firms (see definition below)
are eight times less likely to say their HR technology is troublesome,
a simple but effective metric for evaluating software usability.6 And
many studies in recent years put usability at or near the top of the list
of requirements when evaluating vendor options, which means it’s no
longer a ‘nice to have’ but is a ‘must have’ for enterprise HR software
buyers.
6 ARTIFICIAL INTELLIGENCE FOR HR

DEFINING ‘HIGH PERFORMING’

One of our standard research practices is to attempt to separate out high-

performing companies from the rest by asking employers in our surveys to
explain changes in key performance indicators (KPIs) over time. Those KPIs are
then used to signify whether an employer is a ‘high performer’ or not. The specific
metrics we use are revenue, employee retention and employee engagement, as
they are linked in a large body of research to each area of the business.
For instance, revenue is related to business health, but it also signifies a
positive customer experience because satisfied customers spend more, are
more likely to purchase again, and are likely to refer other customers.
Employee retention signifies a positive employee experience across a range
of areas, and retention directly impacts company profitability. Engagement
not only signifies a positive workplace experience, but other business areas
as well. Several studies correlate engagement with profit, innovation,
customer satisfaction and more. The three elements of revenue, retention
and engagement paint a picture of organizational health and give us a
valuable measure by which to cut our research data.
It is common to see drastic differences in employer practices based on
whether the firm is a high performer or not by our standards. Consider
these three examples:
●
In our ‘Business value of HR technology’ study, the research shows
high-performing firms are eight times less likely to say their HR
technology is troublesome.7
●
Our ‘Performance management, engagement, and business results’
study pointed out the talent practices that high performers were more
likely to implement: frequent goal-setting, recognition for performance,
in-the-moment feedback, developmental coaching and peer feedback.
Low performers were more likely to prioritize annual reviews and trying
to develop employee weaknesses.8
●●
Our latest ‘Learning content strategy’ study showed that high-performing
firms were more than twice as likely to be using a cohesive strategy to
guide learning content development and delivery, and they were also
twice as likely to be measuring learning impact and outcomes.9

While these aren’t the only ways to measure success, they provide a powerful
lens through which to analyse practices and approaches that high-
performing companies have in common. Within the context of this book, any
mentions of high-performing practices are based on these criteria.
 A SNAPSHOT OF HR TODAY 7

Researching and understanding HR today

The research practice our team leads has uncovered a wide variety of
insights in the last few years, and they paint a picture of HR that
requires more sophisticated practitioners and tools than those of
yesterday. Across the spectrum of human capital management, from
talent acquisition and employee development to talent mobility and
engagement, greater demands are being placed on the human resources
function to deliver tangible, actionable business results. Some exam-
ples of shifting trends are included below to help set the stage for
discussions to come around specific human capital management func-
tional areas.

Talent acquisition
Today’s hiring climate is a candidate’s market, which was confirmed
in a recent research interview with the founder of a talent acquisition
technology firm. In the discussion the founder explained that he is
seeing incredible pressures not just in normally challenging fields like
software engineering or nursing, but in other areas as well, such as
sales or skilled trades. His belief is that candidates have more power
than ever to demand what they might want in a role, and those
demands change how the company recruits and interacts with the
candidate population. The reason he founded his firm was because
the entire online job search process is disappointing and frustrating
for candidates. His team’s research shows that nearly three-quarters
of people would agree that online job search is a frustrating process.
This is driving a change in technology and company behaviours to
deliver a better candidate experience.
Candidates also want more interaction in the application process –
our research tells us that when candidates apply for a job, they want
a chance to really show what they’re made of. While it’s often been
thought that elements of the hiring process like assessments were
disliked by applicants, the data actually says that they like those
aspects but only if they actually make them more competitive for the
job. In other words, we don’t want a generic psychometric test, we’d
much prefer a job simulation that allows us to showcase relevant
skills and knowledge.
8 ARTIFICIAL INTELLIGENCE FOR HR

At the end of the day, candidates are looking for a more personal-
ized experience, and yet recruiters and employers are struggling to
keep pace with hiring needs on a global scale. The Lighthouse
Research ‘Talent acquisition priorities’ study has consistently found
that talent acquisition leaders are focused heavily on improving their
relationship with the business and improving their practice, but the
primary areas of hands-on recruiting that they want to fix are
onboarding and sourcing.10 In virtually every research study on talent
acquisition I’ve conducted or seen in the last seven years, sourcing
technical talent always rises to the top of the list. Another research
project pointed out just how challenging it is to scale globally. While
today’s workplace is more globalized than ever before, employers still
struggle with the leap to international operations. Seven out of ten
employers with a global presence are not fully confident in their
knowledge of foreign compliance requirements, and four in ten
employers are spending more than four hours per month per employee
to onboard, pay and manage global workers.11 This hampers not
only HR’s productivity but its ability to influence and impact perfor-
mance across the enterprise.

Talent mobility
Employees want to own not just their jobs but their careers as well.
The first research project I ever completed focused on entry-level HR
professionals and their primary needs, and one of those turned out to
be a very clear career path. Fast-forward ten years and the data I’m
seeing now from other sources validates those findings, in that
employees and job applicants are hoping not just for a job but for a
longer-term career track that they can plan for. The intriguing inter-
section of this, however, is that employees are perceived as staying at
jobs for less time than in years past. This perception continues to
drive decisions, spending and resource investment, despite the fact
that it has no basis in reality. The United States Bureau of Labor
Statistics reported in 1983 that the average workplace tenure was 4.4
years.12 If we believe the current narrative, the records should show a
sharp decline in this tenure number because of job-hopping and
 A SNAPSHOT OF HR TODAY 9

unrest in the employee population. Yet data from 2016 shows the
average tenure to be 4.2 years, a difference of approximately two
months – hardly worth mentioning as a major trend in the employ-
ment landscape.13
Employees want career guidance, but they also want skills growth.
Our talent mobility research has uncovered a variety of companies
that have somewhat radical approaches to growing and managing
their workforce, from allowing employees to pick up and move across
the organizational chart to a manager they feel is more likely to
support their developmental needs, to firms that offer stretch assign-
ments with very clear goals and outcomes as a way to create value for
individuals, teams and the business. The challenge is that while many
employers want to participate in these types of programmes, they are
not sure how to make them a reality. The leap from theory to applica-
tion can be considerably difficult because of potential conflicts with
culture and organizational norms. The Lighthouse Research talent
mobility profit chain (Figure 1.1) shows a high-level overview of how
these talent practices relate to the results of the business.
This need for growth is partially driven by the widely recognized
skills gap. Research from Udemy, a learning technology platform,
shows that 80 per cent of Americans believe there is a skills gap, and
this number is fairly consistent globally.14 The participants in the
research also said that the skills areas they most need to focus on
(from most to least important) include technology skills, leadership
and management, productivity skills, interpersonal work skills and

FIGURE 1.1 Lighthouse Research & Advisory talent mobility profit chain

Research shows that talent mobility and Engagement has been shown to be
development opportunities are drivers of linked to key business outcomes,
employee engagement. such as:

• revenue
• retention
• innovation
• customer satisfaction
• profitability

SOURCE lhra.io
10 ARTIFICIAL INTELLIGENCE FOR HR

soft skills. In our own research we see that among North American
and European workers, 60 per cent say that they have to figure out
their skills on their own without input from their leaders. While
learners put the primary burden on themselves for skill development,
they also acknowledge the importance of the respective roles of
managers, company leadership and HR/talent professionals to make
it a priority.

Learning agility
Taking the previous discussion into the employer’s hands, busi-
nesses need to be looking at how to train and teach workers in a
way that improves the firm’s overall agility in the marketplace.
Training is about more than just safety – it’s a way to continuously
invest in the success of the organization over time. One area we’ll
focus on shortly is the concept of the ‘employee experience’, which
translates to the learner experience in this context. In the course of
research for a project on informal learning tools and measurement,
I spoke with a global learning and development leader about the
idea of the learner experience. His response surprised me at the
time, but in hindsight it is very appropriate. He said that the whole
idea of creating a positive, friendly experience for learners was a
complete waste of resources and time if the intent was only to create
that type of experience. However, if creating that positive experi-
ence led to other benefits, such as increased learner engagement, or
retention of materials or performance, then it was a valuable process
in which he’d be more than happy to invest his training budget. In
other words, the experience can’t be the outcome – it has to lead to
a worthwhile outcome.
At the same time, employers are faced with a decision on how and
where to invest their learning budgets in a high-impact manner.
Technology plays a part, and we’ll dig into the AI applications for
learning and development later in the book, but core components
like training and learning content are also important to how learning
gets accomplished within an organization. Our research shows that
high-performing employers are more likely to look at learning content
 A SNAPSHOT OF HR TODAY 11

for the engagement value it creates. For those companies, value comes
in the form of voluntary consumption and adoption rates, not just
considering quantity of content development or course completions
as the ultimate goal of the learning and development function. It’s
about quality more than quantity in this case. Low performers are
seven times more likely to say their learning content does not engage
learners or is merely transactional in nature.15 This may be due to the
starting and ending points: the strategy around creation and meas-
urement. Many companies are not using a comprehensive strategy to
guide their learning objectives, yet we all know that it’s impossible to
hit a target if we don’t first aim at it. Employers that try to use learn-
ing as a differentiator for creating a more valuable workforce are
doomed to fail if their approach is purely ad hoc. Filling skills gaps
on a first-come, first-served basis means that learning and develop-
ment (L&D) teams will always struggle to keep up with demand.
Taking a more strategic approach and prioritizing content develop-
ment and delivery based on business objectives is more likely to lead
to high-impact learning programmes, which is why high performers
are twice as likely to have a strategy leading their learning invest-
ments and activities (Figure 1.2).

FIGURE 1.2 Approach to learning content strategy

60%

50%

40%

30%

20%

10%

0%
Beginning: We have Developing: We have Leading: We have a
little to no consistent some strategic strategy in place and
strategy driving elements but no are using it to
our learning strong governance of drive our content
content our learning creation
content and delivery
High-performing Low-performing

SOURCE Lighthouse Research & Advisory, 2017 Learning Content Strategy Study
12 ARTIFICIAL INTELLIGENCE FOR HR

Racial justice
In the last few years there has been an incredible surge in the focus and
prioritization of racial justice within the employment community. It
has evolved from the deeply rooted diversity focus, where employers
primarily looked at numbers and quotas. Next came inclusion, and
diverse individuals asked not just to be hired but to be valued for what
they brought. Now we’ve moved into a larger focus on justice and
equity, ensuring that there is support for the diverse needs of the work-
force across a range of practices. For instance, in 2019 I had the
pleasure of spending time with the team at the healthcare giant Kaiser
Permanente during their annual internal event celebrating diversity,
equity and inclusion (DEI). The company’s chief diversity officer was
incredibly proud to talk about and highlight the importance of creat-
ing a culture at Kaiser Permanente where everyone, regardless of race,
sexual identity or other identifier, felt like they truly belonged at the
firm.
This kind of commitment to being an employer of choice for diverse
individuals is what I’m seeing from employers around the globe.
During a recent interview with Dominique Brewer, a talent leader at
Takeda Pharmaceuticals, she explained that the firm was renewing its
commitment to justice and equity in some innovative ways.16 The firm
wasn’t just trying to hire diverse individuals – leaders were also look-
ing for ways to improve equity outcomes beyond the employee lens.
For instance, they are ensuring that pharmaceutical trials accurately
represent the diverse populations they serve, not just one demographic.
They’re also embedding this equity focus into procurement and other
processes within the business. She said it’s been powerful to see the
internal leaders seize this opportunity and run with it, ultimately
creating a more equitable business and impact on the world.
And this overview could easily go on, delving into any number of
areas from analytics and onboarding to workforce management and
engagement. The world is becoming more complex, not less, and
employers need to seize the current opportunity where there is more
data and insight than ever before on hand to help solve the problems
that exist. One area that offers promise in solving those problems is
the employee experience, because instead of trying to treat smaller
 A SNAPSHOT OF HR TODAY 13

segments of the population, employers can craft a more holistic

­strategy for improving the workplace.

The power of the employee experience

One of the most talked-about topics in all of human capital manage-
ment in recent years has been the concept of the employee experience.
Books, articles and presentations all tie back to this idea of creating
a series of workplace experiences that create immense value for our
employees. Employee engagement is a similar topic, but it’s commonly
seen as a subset of the broader employee experience, and engagement
scores are used as one metric for determining the overall value of the
employee experience. Any practitioner can tell you the statistics:
employee engagement isn’t improving, and hasn’t in years. Gallup’s
data says that about one in three employees are engaged and that
number hasn’t changed significantly in more than ten years.17 That’s
a key reason so many industry leaders have seen the employee experi-
ence as a potential solution to the perpetual engagement problem.
Aptitude Research Partners’ 2016 data shows that employers see
these concepts as very similar, with 83 per cent of companies agreeing
that ‘improving the employee experience’ and ‘developing a formal
engagement strategy’ are their top priorities.18 Consider these reasons
why employee experience is a must-have item on the agenda for those
HR and business leaders examining workplace issues and solutions:

1 Congruent customer and employee experiences

How can we expect our customers to deliver superior service and
customer experiences when their own experience as an employee is
lacklustre or even worse? I’ve long said that the customer experience
will never exceed the employee experience. Well, what I’ve actually
said is, ‘Employees will never treat customers better than their
management treats them’, but it’s one and the same. The experiences
will be congruent, or similar. That means companies that live and
die by customer satisfaction scores need to start not with new
customer bonuses or other gimmicks but with a positive employee
experience in order to see the actual outcomes they are looking for.
14 ARTIFICIAL INTELLIGENCE FOR HR

2 The employee experience is not the employee life cycle

One of the issues with the initial attempt to grasp the concept of
the employee experience is to put it in the context of the employee
life cycle. Don’t think that understanding the mechanics of
onboarding and performance management means that you have a
great employee experience. The experience, or how someone feels,
is part of the life cycle, but it’s not quite the same thing. Instead
look at the candidate- or employee-centric nature of your processes
and see to what extent they support, encourage and engage your
workforce. That’s how you can get a true sense of the value you are
creating through your own employee experience.
3 Tell me about your employees first
If I walked up to you right now and asked about your company,
how would you begin the conversation? What would you start
with? Your products? Your mission? Your customers? What about
your employees – would they even make it into the discussion? I
once asked this question of several dozen business leaders as an
experiment, and not a single one of them responded to me with a
description of their employees or the kinds of people they hire. It’s
so common to think about this question in the context of customers
or services when in reality it’s our employees that make us successful.
Start the discussion with employees as your central topic and go
from there. It will change the perspective of those around you.
4 Expectations rule the day
A big part of why employees have bad experiences in the workplace
is because of expectations. Have you ever had high expectations
for a pay rise, performance discussion or meeting, only to walk
away feeling disappointed? Or maybe your first few days on the
job are a complete waste of time sitting around waiting for
resources, despite being treated very positively in the hiring
process? The theory of expectancy plays into motivations and how
we feel about choices we make. If you want to deliver a positive
experience, make sure you give people a warning ahead of time so
their expectation gap (what they expect and what you ultimately
deliver) isn’t as large.
 A SNAPSHOT OF HR TODAY 15

5 Companies don’t really exist, people do

The trouble with leaders in many organizations is that they view the
company as ‘The Company’, an autonomous entity that doesn’t
need to be understood or afforded respect. In this worldview,
decisions are made as if employees are replacement parts, and we
don’t have to worry about the feelings of replaceable parts. This is
exemplified by the way one company manages its workforce. A
popular ridesharing company actually uses an algorithm to assign
shifts and communicate with workers, which has led to drivers
cheating the system in order to improve their financial outcomes.
This story is highlighted more fully later in the book, but the key
element to remember is that companies can’t assume that people are
emotionless automatons, because they aren’t. People, not ‘The
Company’, get stuff done. People are the face of the firm, not a logo,
billboard or slogan. Remember that.
6 Success results from design thinking
The concept of design thinking centres on this: efforts are spent
not just on solving problems but on creating solutions with the end
result in mind. In this case, how can we create solutions that focus
not on the organization or solely on the customer but on the
employee experience? Instead of thinking about how to fix a
problem specifically, the focus is on becoming something radically
different. For many of us, that’s the direction we need to go to
rectify design flaws in our processes and policies that can actually
hamper our efforts to engage our workers. For instance, in a recent
interview with an HR executive, she explained that the company
had done away with the normal ‘doctor’s note for sick leave’
requirement. She said that when someone got a cold, they would
respond in one of a few ways: sending them to the doctor to get a
piece of paper even when the doctor could not improve their
condition often made them angry or upset; some workers would
continue coming to work even with a cold because they didn’t
want the hassle of scheduling a doctor’s appointment in the middle
of their illness. This is a small example of how HR leaders can look
for ways to design processes and approaches that support employee
needs first and business needs second, because in reality taking
care of employees is how you take care of the business.
16 ARTIFICIAL INTELLIGENCE FOR HR

The new normal: Doing more with less

These and other talent-specific challenges are facing HR leaders
around the world every day, and that’s not all. On top of this we have
the constant barrage from the business to ‘do more with less’. One
CHRO I spoke with recently told me that his company consistently
reduces his budget between 5 per cent and 8 per cent a year, forcing
him and his team to make hard choices on a regular basis. This is a
reality, and has been for some time, for many companies. Of the vari-
ous areas of the business, this concept seems to hit human resources
harder than many others. For example: a one to one hundred ratio is
commonly seen as the ‘standard’ ratio of HR staff to employees, yet
if that ratio shifts to one to one thousand, there is virtually no time
for that HR staff member to participate in anything even remotely
strategic. All efforts are spent on day-to-day activities, shuffling paper
and other administrative work. Even if none of these other shifts and
challenges were present, this one would remain as an incredible
barrier to HR leaders being able to deliver high-value service to both
the employer and employee populations.
When we combine the leanness of today’s HR teams with the need
to create more personalized experiences for workers, we have a perfect
storm of demands that simply can’t be met in the traditional manner.
All this is not meant to show that HR can’t overcome these challenges.
With the right tools, we absolutely can. Consider the consumer exam-
ples of how technology enables highly personalized experiences at
scale. ESPN re-launched its website properties, allowing users to spec-
ify which sports or teams they are interested in. This gives each
individual a highly customized experience from the exact moment
they visit the website. Our mobile devices allow us to set up specific
applications and options in ways that make us more productive, and
no two people have the same exact settings and preferences because
no two people are exactly the same. While simple, these examples
provide a hint of the value that technology can provide in the work-
place as well by creating highly personalized experiences regardless of
whether the audience is a candidate, a line-level employee or a busi-
ness leader. For instance, one example we’ll explore in a coming
 A SNAPSHOT OF HR TODAY 17

chapter shows how a recruiting technology provider created an inter-

face that allows candidates to have conversations with a humanized
chatbot, focusing in on their most relevant experience and interests
just as a human interviewer might in the screening process. This isn’t
just about automating that workload, which has its own inherent
value – it’s also about creating that personalized feeling for those
that interact with these sorts of systems so that they feel appreciated
and understood.
Leveraging new technologies can bring about myriad improve-
ments in employer operations, and the core of many new technologies
is some variety of artificial intelligence. While AI may conjure up
images of robots and movie plots, the type of AI discussed here is
more benign. It’s less physical and more technical. It’s less about
general purpose AI, which doesn’t exist outside of a science fiction
movie, and more about leveraging automation and other capabilities
of AI for highly specialized data sets, predictions and outcomes. The
discussion of artificial intelligence today often centres around two key
areas that we’ll explore more deeply in the coming chapters: automa-
tion and augmentation. The intent is to truly do more with less, freeing
up our high-value HR talent to pursue more strategic activities and
letting the algorithms do the repetitive heavy lifting and analysis.

EXERCISE: THINKING BEYOND TODAY

If you’re like most business leaders, today has been busy. It’s been
challenging. You’ve spent some time on things that may or may not have
been on your formal ‘to do’ list, which might mean you’re farther behind
today than you were when you started. It’s important occasionally to rise
above this general busyness and think beyond the day. For the moment,
put yourself in the place of an entrepreneur running your own organization
or department. Think about the resources at your disposal, the tasks you
have to get done, and the value of your time. Oftentimes employees, even
executives, have trouble thinking about the specific value of their time, but
entrepreneurs are often better about realizing what is a core focus area and
what needs to be outsourced or delegated. With that in mind, if you had an
additional hour in your day, how would you spend it? What would you
18 ARTIFICIAL INTELLIGENCE FOR HR

spend it on, or, more importantly, what would you invest it in? Take a
moment and write down two or three things you would do with that extra
time. Would you brush up on professional development needs? Would you
refine some of your processes around gathering and reporting metrics?
Maybe you’d do personal phone calls to more candidates instead of sending
mass emails? Whatever the case, each of us has different priorities, so
consider yours for a moment.
A more personal example of this is a personal budget for someone with
irregular income. It’s challenging to budget month to month with an irregular
income, so financial planners often recommend an irregular income planning
process that lists priorities from most to least important. As money comes into
the household, individuals start applying it to the top of the list and moving
down the priorities. At the top of the list might be groceries or utilities, and
near the bottom would be discretionary spending on areas you care about.
This concept is the same as our discussion on how you’d spend additional time
but helps to make it more concrete because there is a prioritized list that
governs how additional resources are applied as they become available.
This exercise is important because the promise of artificial intelligence
tools is, at its core, more time. By automating a part of HR and having an
algorithm or machine handle the work, that frees us up to do other things. The
problem with this is Parkinson’s Law. Parkinson’s Law states that our work task
will expand to fill the time available. In reality, when we have an innovation
that ‘adds’ time to our day by reducing some administrative requirement, the
rest of our work creeps into that available time in an unobtrusive way,
eliminating those savings if we’re not careful. Time creep happens based on
urgency or some other criteria instead of coming from a strategic look at the
high-value actions that generate the most value for the person, the team or
the company. One final example of this: if an HR leader earns $50 an hour and
a new tool saves that person an hour a week, the benefit to the company isn’t
just the $50 per week in saved time – it should also include the new, higher-
value activity the HR leader is pursuing with that freed-up time.
The coming chapters will include case studies that show greater
automation and time savings. They will tell about revolutionary
technologies and tools that promise to improve the workplace. But if we
don’t think critically about how we’ll use that additional time saved from
automation, then we’re leaving a large portion of the value on the table. As
you proceed through the book, think hard about how these advancements
might apply to your own organization and how you might use those
additional hours to improve your own standing, credibility and value.
 A SNAPSHOT OF HR TODAY 19

Using this book

If it hasn’t happened already, one day an opportunity will present
itself for you to explain the impact that artificial intelligence will have
on the HR function. It’s my hope that this book will help you to
answer that question both on a macro scale (how AI impacts the HR
profession and the competencies necessary for success) and the micro
scale (how this impacts your company and your HR team).
What good is a tool if it’s presented in theory only? This book will
not only help you understand the concept of AI as a technology, but
more importantly you will understand the use cases and opportuni-
ties for HR to adopt these tools and systems to support our
ever-present quest for improving business outcomes through better
people practices.
The best way to accomplish this is not by looking at HR as a giant
mishmash of practices and processes. That sort of viewpoint is exactly
why general purpose AI is more of a buzzword than an actual appli-
cation of practical value today. Instead, we will examine HR through
discrete practice areas such as recruiting and talent acquisition, learn-
ing and development, talent management and more. This book is
designed to help you become a better professional, pure and simple.
If you’re hoping to get a more general idea of how AI works and how
it is penetrating every part of our lives, Chapter 2 will be beneficial to
you. If you are more interested in a specific use case, there are chap-
ters dedicated to answering your questions around items such as
payroll (Chapter 4), recruiting (Chapter 5) or learning (Chapter 6).
In addition, there are challenges with these types of technology,
just like any other. With that in mind, this analysis will also examine
some of the more common challenges with using artificial intelli-
gence, such as system bias. We will also focus on the competencies
that matter most for an HR leader in a digital world, requiring a
mixture of human ingenuity and creativity with the scalability and
insight that machines can provide. This balance has always been a
challenge since the earliest days of automation, but in a world where
knowledge work rules and computers are doing an increasing amount
of knowledge work, the discussion is more pertinent than ever.
20 ARTIFICIAL INTELLIGENCE FOR HR

KEY POINTS
●●
Today’s world of HR is more complex than ever before, and a disruptive
environment requires companies to be more agile in order to respond to
market demands.
●●
Talent acquisition, learning, workforce management and other core HR
practice areas are all affected by changes in candidate and employee
preferences, which places even greater demand on today’s HR function.
●●
Employee experience is more than a buzzword – it’s a way to
encapsulate the entire sum of experiences that employees perceive from
the first application through to the last touch of offboarding. Employers
must ask themselves whether those moments, either in whole or in part,
are generating positive experiences for their candidates and employees.
●●
New artificial intelligence technologies that automate and augment the
workforce could be the key to solving some of the thorny issues and
increased demands for HR to accomplish more with fewer resources.

Notes
1 IBM. Redefining Talent: Insights from the global C-suite study – The CHRO
perspective, 2016. www.ibm.com/downloads/cas/Z8RWPADA (archived at
https://perma.cc/VR5N-4HUL)
2 M Ramsey and D MacMillan. Carnegie Mellon reels after Uber lures away
researchers, The Wall Street Journal, 31 May 2015. www.wsj.com/articles/
is-uber-a-friend-or-foe-of-carnegie-mellon-in-robotics-1433084582 (archived
at https://perma.cc/6EHC-J5E6)
3 P Gillespie. The rapidly growing gig economy is still super small, CNN
Business, 6 May 2016. http://money.cnn.com/2016/05/06/news/economy/
gig-economy-princeton-krueger-tiny/ (archived at https://perma.cc/2TND-
3S6Z)
4 A E M Hess. The 10 largest employers in America, USA Today, 22 August
2013. www.usatoday.com/story/money/business/2013/08/22/ten-largest-
employers/2680249/ (archived at https://perma.cc/F4PT-K9CH)
5 B Eubanks. Is there a relationship between HR technology and organizational
agility? Lighthouse Research & Advisory, 17 September 2020. https://lhra.io/
blog/is-there-a-relationship-between-hr-technology-and-organizational-agility-
new-research/ (archived at https://perma.cc/BGD5-KK2B)
 A SNAPSHOT OF HR TODAY 21

6 B Eubanks. The business value of HR technology, Lighthouse Research &

Advisory, 1 August 2017. http://lhra.io/blog/business-value-hr-technology-
research-preview/ (archived at https://perma.cc/YKB7-355D)
7 https://lhra.io/blog/business-value-hr-technology-research-preview (archived at
https://perma.cc/GX79-KHYE)
8 https://lhra.io/blog/how-high-performing-companies-manage-employee-
performance-new-research (archived at https://perma.cc/Q55J-57EG)
9 https://lhra.io/blog/new-research-radically-rethinking-learning-content-
strategy-free-report (archived at https://perma.cc/D2CL-KWRP)
10 B Eubanks. Today’s top talent acquisition priorities, Lighthouse Research &
Advisory, 9 October 2017. http://lhra.io/blog/todays-top-talent-acquisition-
priorities-free-ebook/ (archived at https://perma.cc/BWS4-3F5U)
11 B Eubanks. Global HR practices: Compliance, growth, and productivity,
Lighthouse Research & Advisory, 26 June 2017. http://lhra.io/blog/global-hr-
practices-compliance-growth-productivity-new-research/ (archived at https://
perma.cc/36QA-6EA7)
12 E Sehgal. Occupational mobility and job tenure in 1983, Monthly Labor
Review, 1984. www.bls.gov/opub/mlr/1984/10/art2full.pdf (archived at
https://perma.cc/Z75P-WFX9)
13 US Bureau of Labor Statistics, Employee tenure summary, 22 September 2020.
www.bls.gov/news.release/tenure.nr0.htm (archived at https://perma.cc/
N5AT-DSHC)
14 Udemy. 2017 skills gap report, 2017. https://research.udemy.com/research_
report/2017-skills-gap-report/ (archived at https://perma.cc/LWN8-ZVKU)
15 B Eubanks. Radically rethinking your learning content strategy, Lighthouse
Research & Advisory, 14 November 2017. http://lhra.io/blog/new-research-
radically-rethinking-learning-content-strategy-free-report/ (archived at https://
perma.cc/N4C9-ZFAL)
16 B Eubanks. How Takeda radically transformed its HR operating model, We’re
Only Human [podcast], 15 October 2020. https://onlyhumanshow.com/e/
how-takeda-radically-transformed-its-hr-operating-model-on-were-only-
human/ (archived at https://perma.cc/X2ER-R47B)
17 A Mann and J Harter. The worldwide employee engagement crisis, Gallup, 7
January 2016. http://news.gallup.com/businessjournal/188033/worldwide-
employee-engagement-crisis.aspx (archived at https://perma.cc/4WAX-PB9S)
18 B Eubanks. The value of the employee experience, Lighthouse Research &
Advisory, 5 December 2017. http://lhra.io/blog/value-employee-experience-
free-ebook/ (archived at https://perma.cc/M7G2-HJGS)
22

THIS PAGE IS INTENTIONALLY LEFT BLANK

 23

02

The basics of artificial intelligence

Note: The intended goal for this chapter is to be as short as possible. The
priority is to allow you to move on to the more practical elements of this
book while still being long enough to help you understand how the data
analysis and artificial intelligence aspects of these new technologies
operate. For those that have a technology background and want additional
information I will share a few resources at the end of this chapter for
further reading, but for those that are hoping to see how this truly impacts
the HR profession, this will carry you right into Chapter 3, where we really
start delving into specific examples of technologies and broad use cases for
human capital management.

Between 14 million and 21 million lives. That’s how many people are
­estimated to have been saved by the code-breaking genius of Alan
Turing and his team. Known as ‘the father of modern computing’,
Alan Turing was a brilliant mathematician who dedicated his time
during World War Two to breaking codes and ciphers. During the
conflict, one of the most challenging tasks undertaken by the British
military was to break the codes used by the Axis powers in order to
understand troop movements, attack plans and other important
information. Perhaps the most difficult coding technology in the
world at the time was known as Enigma, the system used by the
Germans. After some time and an extensive amount of effort, a team
of British code breakers were able to crack Enigma. In an analysis of
24 ARTIFICIAL INTELLIGENCE FOR HR

the data, researchers estimate that cracking the Enigma code using a
system designed by Turing led to shortening the war by several years,
saving millions of lives in the process.1
What does this have to do with artificial intelligence? One of
Turing’s other contributions based on his mathematical research was
the concept of a ‘thinking machine’.2 Turing envisioned a machine
that could think like a human and even proposed an experiment,
called a Turing Test, to determine whether a machine could pass as a
human. The Turing Test is a process whereby an interrogator interacts
with a person and a machine via an electronic interface and then has
to guess which is the human and which is the machine. In order to
pass the Turing Test, a machine has to be able to passably simulate a
human level of intelligence in the short duration of the examination.
In spite of the tragic loss of Alan Turing before his time, his name is
regularly referenced today with regard to artificial intelligence because
of his significant early contributions to the field of research.
While the stakes are lower than those associated with code break-
ing during wartime, the technologies we use today in the workplace
are still incredibly important. Over the years I have seen hundreds of
technologies, analysed their functionality and advised businesses on
how to develop their products and market approaches. The reason
they seek this advice is partly because I bring a practitioner’s view-
point to the conversation, helping them to think about product
development from their buyer’s mindset. When I’m looking at a piece
of technology, I think about my time working as an HR executive
and ask myself the question, ‘Would this have made my life easier in
a practical way, or is it more flash than substance?’ It’s a challenging
lens because not all software meets that threshold of ‘practical value’,
but it’s important if someone wants to serve an audience of HR and
business leaders to be clear about the value the system can bring to
the stakeholders. The other reason vendors seek advice is because I
have researched and analysed trends in the HR and learning technol-
ogy industries. While there’s no way to intimately know and stay
connected with all the providers in the industry, doing several brief-
ings and product demos every week keeps me fresh on the latest
advancements and directions of key providers. This insight into
 THE BASICS OF ARTIFICIAL INTELLIGENCE 25

t­echnology offerings matters, since a product conversation rarely

occurs these days without a mention of machine learning, algorithms
or other AI-based concepts. It’s possible you’ve also seen a demo,
heard a sales pitch or read about a vendor’s technology that relies on
artificial intelligence in order to operate. This chapter will help to
give you the essential grounding in AI technologies in plain terms,
making the conversation as practical as possible and helping you to
cut through the marketing language and understand just how a
particular system leverages AI to operate. Or, as often happens, you’ll
be able to see through the hype and understand when someone might
be overinflating the capabilities of their particular system.
I’ve already established my intent to make this book highly practi-
cal and actionable for you, which requires a knowledge of application,
not just theory. The best way to understand how we arrived at this
intersection of artificial intelligence and the human resources profes-
sion is to take a step back and look at the logical progression of
things over time. AI tools such as machine learning require immense
amounts of data – millions of data points – to be successful. Not
surprisingly, the business buzzword of recent years has been ‘big
data’, which is a foundational underpinning of a system powered by
artificial intelligence. Without the improvements in data creation and
capture in recent history, there would be no opportunity to utilize AI
to drive predictive models. By gathering the necessary amount of
data, we can then predict trends, outcomes and more by understand-
ing the variables. Additionally, research from Sierra-Cedar’s report
explores a category of firms called talent-driven organizations, those
that use analytics to solve key talent challenges around engagement,
retention and identification of top talent. The research finds that
talent-driven organizations are much more likely to have machine
learning-driven business intelligence and HR analytics tools in place.
These types of firms are 166 per cent more likely to be using these
systems and 68 per cent more likely to be evaluating new systems to
solve the problems associated with talent analytics.3 That data point
helps to demonstrate why this discussion matters, so consider the
following layperson’s definition of predictive analytics to help lay the
groundwork for the exploration of AI that is to come.
26 ARTIFICIAL INTELLIGENCE FOR HR

A layperson’s guide to predictive HR analytics

One of the challenges of talking about predictive analytics, big data
and other similar concepts is that it quickly becomes both complex
and abstract, causing the general audience to tune out of the conver-
sation. This is especially relevant for HR leaders, where competencies
in these areas are somewhat weak in general. By the way, that’s not
an indictment on HR leaders, it’s just a statement of the facts. That’s
not to say they aren’t great leaders or don’t have the skills and experi-
ence necessary to work within the human capital management field.
It’s just that the need to gather, analyse and predict outcomes using
data has not been a required skill set for as long as some other types
of skills and competencies.
I had a conversation with a technology firm founder who talked
about some of the work he had done supporting NASA and other
government technology projects over the years. He said that an inter-
esting shift happened over time. In earlier decades, the focus was
mainly on hardware, and software was a very minor component.
Systems were built out of ‘dumb’ parts and had a very limited amount
of computers and other technology to support their operation. Over
time, the software element increased exponentially both in power
and in impact as technology improved. There was a time somewhere
in the middle of this ongoing transition where some hardware engi-
neers had to ‘become’ software engineering subject matter experts.
While this group of individuals might not have had formal training
or education in software systems, the workers had the right under-
standing of the programs, customers and objectives that a newly
trained software engineer just out of college wouldn’t understand.
Their limitations on the technical side didn’t hamper their ability to
add value to the rest of the process.
I see the same thing happening today within HR. In recent years
we’ve felt the need to turn HR leaders into analytics experts, even
though that might not be their core skill set. Whether you are an
expert or brand new to the topic of HR analytics, I want to help you
understand the impact of analytics and how they work. My goal is to
transform you not into an expert able to carry out analyses and
 THE BASICS OF ARTIFICIAL INTELLIGENCE 27

TABLE 2.1 Types of talent analytics

Analytics type Purpose Example

Descriptive Tells current state Retention rate
Predictive Correlates data points to Personalized retention
predict changes predictions
Prescriptive Suggests potential solutions Personalized retention
to predicted changes suggestions

perform statistical modelling, but into a more conscious and educated

participant when it comes to those analytics and data conversations
I mentioned previously. Let’s start with a quick overview of the three
types of analytics (Table 2.1) before jumping into the conversation
around the predictive analytics that can be generated by many of the
machine learning tools today.

What are predictive analytics?

In the book Predictive HR Analytics, the authors talk about three
specific ways to discuss this topic:
●●
identifying predictors and causal factors
●●
predictive modelling
●●
predicting behaviours4

Again, the goal is to present this in layperson terms, so let’s dive in.

CONCEPT 1: IDENTIFYING PREDICTORS AND CAUSAL FACTORS

In this instance, we are trying to find out what variables are linked to
each other. Data can be tied together by correlation or causation.
Correlation simply means that there seems to be a relationship in the
data (for example, people seem to carry umbrellas more often when rain
is predicted). Causation is something else entirely (for example, we know
that carrying an umbrella does not cause it to rain more frequently).
If we can identify what variables feed into others, then we can use
those drivers or levers to create the results we need. For instance, if
28 ARTIFICIAL INTELLIGENCE FOR HR

we can link increased training to higher sales, then that would seem
to be a causal factor. Testing would need to be done to determine the
extent of the linkage, but you get the picture. The first, most basic
step for prediction is finding those predictors. Then we build on top
of that foundation.

CONCEPT 2: PREDICTIVE MODELLING

This takes the conversation a step further. Let me clarify really
quickly – we are talking here about leading variables. When a leading
variable changes, it affects other elements down the line. Think about
it like an assembly line – if we change something on the front end, the
rest of the process is affected dramatically. If we mess up on the front
end, then the rest of the process is affected in a different way. That’s
how we use leading variables within the predictive conversation.
Once we have identified the predictor variables from the previous
section, we step up a notch and start trying to predict what happens if
we change one of our predictor variables. For instance, if we continue
with the training/sales link mentioned above, the goal might be to try
to see if doubling training also doubles sales. Or, in another context,
maybe we find out that there is a link between manager communica-
tion and employee engagement. Then we start trying to model whether
increasing or decreasing manager communications affects engage-
ment and to what extent. The point here is to focus on the driver
variables and determine what happens if we start changing them
around. How do they impact the final result? What changes occur?

CONCEPT 3: PREDICTING BEHAVIOURS

The final and most complex piece is determining what happens if we
apply our model to new data or populations. In other words, can we
predict how people will respond to certain variables?
Let’s say we have data on employee turnover that is related to a
variety of factors, including manager check-ins, performance evalua-
tion scores and tenure. By mapping all of those variables for existing
employees, we can create a model that will allow us to predict future
behaviours. For instance, if the data shows that fewer manager check-
ins, shorter tenure and lower performance scores indicate someone is
 THE BASICS OF ARTIFICIAL INTELLIGENCE 29

more likely to leave, we can put that person in a ‘high risk’ bucket.
The person is more likely to leave than someone that doesn’t have
those types of factors working against them.
Another example could be around recruiting. Let’s say that inter-
view teams of five or more people have been proven to screen better
employees with higher performance than teams of four or fewer
interviewers. With the right data behind the model that proves this
relationship, we can make the case that having five interviewers,
while a fairly significant investment of time, is a good trade-off for
improved employee performance for the duration of employment.
The next step, of course, would be to quantify the improved perfor-
mance and how much value it ultimately drives for the organization.
As you can see, there’s incredible power in leveraging large amounts
of data – a core component of how AI tools function. Now, let’s
examine the level of artificial intelligence development and market
penetration as it stands in today’s market.

Artificial intelligence components

AI is a topic that is bandied about more often today than ever before,
yet it’s often completely misunderstood. While robots and systems
that act without regular human inputs do indeed exist, many newer
tools are based on more rudimentary elements of AI, not self-aware,
intelligent machines that learn from every interaction. The term ‘arti-
ficial intelligence’ is actually an overarching category with several
more targeted terms falling under that heading, including:
●●
machine learning
●●
natural language processing
●●
deep learning
●●
neural networks

Each of these individual technologies is powerful, but when combined

they create opportunities to eliminate wasted time, improve
30 ARTIFICIAL INTELLIGENCE FOR HR

­ roductivity and drive better results. But before we dive into how
p
these technologies work, it is helpful to understand some of the
fundamental components.

The definition of artificial intelligence

If your hopes of what AI can achieve are based on science fiction
movies, then you’ll probably be somewhat disappointed to learn that
many technologies today are not yet artificially intelligent to the
degree we might hope or expect. Even as early as the 1950s, research-
ers expected to have artificially intelligent systems in place by the
year 2000, but this didn’t occur as they had hoped. It’s important to
see artificial intelligence as a spectrum, not as a single destination.
For example, any physics student with a graphing calculator has use
of an advanced piece of technology that automates equations and
processes data faster than the average human. Does that qualify as an
AI system? Systems and tools like this may perform repetitive func-
tions or automate a specific task, but that doesn’t necessarily reflect
the full promise AI has to offer. To establish a definition, let’s consider
what true AI looks like from Lexico, powered by Oxford Dictionaries:

The capacity of a computer to perform operations analogous to learning

and decision making in humans.5

While we could spend time debating the definition of AI and what it

means, I prefer to take the same viewpoint as one of the most influen-
tial and ambitious research efforts around artificial intelligence, the
Stanford One Hundred Year Study on Artificial Intelligence (AI100).
The AI100’s definition of AI points out the fact that a lack of a highly
precise and universally embraced definition of AI is one of the key
reasons for the explosive growth in the field.6 The general sense of
pursuing a human-like level of processing and perception is enough to
drive the field of research and uncover practical applications.
In plainer language, artificially intelligent systems have to not
only analyse activities and predict outcomes – they also need to learn
from those predictions over time. The actual analysis and prediction
is relatively easy. Software companies have been able to do that for
 THE BASICS OF ARTIFICIAL INTELLIGENCE 31

years, and the technology is fairly stable. IBM’s Watson is a great

example of a technology that makes millions of predictions a day on
everything from which employees are going to leave the organiza-
tion all the way to examining medical images to determine whether
patients have cancer. An excellent resource on this topic of predic-
tions is Prediction Machines, a book that explores the economic
impacts of increasingly cheaper prediction capabilities offered by
today’s technology.7 Through an economics lens, if predictions are
cheaper, they should be more plentiful. Lower cost leads to greater
supply, whether that is a supply of fruit, vehicles or even predictions.
If these predictions are more plentiful then we can use them in a
wide variety of applications. This proliferation of predictions is
already a reality today, as you’ll see with some of the applications
we’ll explore in the coming pages.
The challenge for most AI technologies comes in the learning
aspect. Intelligence at a human level is based on constant surveying of
information, learning from that information and adjusting decisions
accordingly. It’s always been that way. In the early days when a cave-
man was eaten by a sabre-toothed tiger, his neighbour watched and
decided (wisely) to avoid those animals in the future. That’s the essen-
tial component of intelligence. In organizational development terms
we call this ‘informal learning’, or learning that occurs in the every-
day course of life and work. In contrast, formal learning is intentional
training delivered with the expectation of improving intelligence or
performance. Informal learning comes naturally to humans and is
something we can do with relative ease, yet it’s more challenging for
computers to grasp the nuances of every situation.
In the book Range, David Epstein talks about the concept of kind
and wicked learning environments.8 A kind learning environment is
similar to golf, where you have a repeatable procedure, very strict
parameters, immediate feedback and an opportunity to adjust your
approach before the next procedure. AI is very well suited for usage
in those kind learning environments because it can pick up and
process the feedback, enabling better decisions on subsequent rounds.
Any process or activity in our lives with the elements of a kind
­learning environment provide fertile ground for AI to be used.
32 ARTIFICIAL INTELLIGENCE FOR HR

A wicked learning environment, on the other hand, may not repeat

regularly and may give little to no discernible feedback to help the
individual adjust for the next scenario. I’m sure we all can agree that
many of the experiences we face day to day are more like a wicked
learning environment than anything else. We make a decision at work
and may not see the results for months. We choose a doctor based on
a very limited amount of information. We give our kids advice based
on our own experiences. This wicked environment is challenging for
us as humans, and it’s doubly so for AI, because the number of vari-
ables in some decisions is seemingly infinite.
An interesting concept that relates to this idea of how machines
learn is Moravec’s paradox.9 The core element of the paradox is that
scientists can teach machines to do difficult or complex tasks like play-
ing chess or solving algebra equations yet they can’t replicate the
capabilities of a toddler to recognize faces, walk across a room or asso-
ciate words with items they see. One possible cause for the paradox
outlined by Hans Moravec is due to evolution. Over time, humans
have learned and mastered critical skills that seem effortless to us, such
as walking or comprehending speech. Moravec posed the idea that the
older and more ingrained the skill, the longer it will take to reverse-
engineer it for purposes of recreating it with software. Strangely, skills
that are more difficult for humans to grasp, such as geometry or calcu-
lus, are more easily programmed into a computer’s capabilities.
If you read any headlines today, you might see some of the more
vocal opponents of AI talking about how the systems and tools in
place are not really ‘intelligent’. There’s actually a name for this
phenomenon: the AI effect.10 In essence, once a machine is able to
demonstrate a human process or capability, onlookers dismiss it as
an advance in computational power or computer modelling, not as
an advance in machine intelligence. When IBM’s computer was able
to beat the world’s best chess player, it wasn’t AI – it was called out
as leveraging simple probabilities, for instance. Noted computer
scientist and mathematician Larry Tesler is known for saying
‘Intelligence is whatever machines haven’t done yet’.11 Once a
machine has done a task, we cease to think of it as intelligence and
instead see it as something else, such as statistical modelling. While it
 THE BASICS OF ARTIFICIAL INTELLIGENCE 33

must be frustrating for AI developers and computer scientists to face

this reality, from a practical standpoint it doesn’t change the results.
I’ll say it again: whether it’s ‘intelligence’ or not doesn’t matter if it
achieves the intended result. Intelligence can be just as much of a
philosophical or theological discussion as a technological one. If a
computer can perform a task that a human could perform, that falls
somewhere on the spectrum of artificial intelligence.
This general pushback on the definition of intelligence is due to
what Tesler believes is an innate human desire to be different and
unique. After all, if we have a computer that can copy what we do as
humans, that makes us less unique in the world, right? However, as
you’ll see throughout this book, and particularly in the chapter on
future-proofing your skills, that is not necessarily true. There are
plenty of capabilities humans have that robots and algorithms can’t
easily match. Think of it this way: when Microsoft turned the full
power of its web services toward translating Wikipedia’s three billion
words across five million articles from English to another language at
its Ignite event in 2016, that translation occurred in less than one
second.12 However, despite that translation and all the raw power at
their disposal, the systems can’t call you on the phone and have a
half-hour conversation with you. There is still a large gap between
the capabilities of today’s technologies and the very human nature of
life and work.
I agree that ‘general AI’ doesn’t really exist today in the form of a
machine that can mimic a wide range of human abilities, but there
are plenty of opportunities to leverage more specific, targeted solu-
tions. While a computer might not be able to quickly learn how to
brush teeth, cook breakfast and drive to work, different systems and
applications might be able to analyse your dental records to find cavi-
ties, examine a photo of your breakfast and estimate the calories, or
offer route alternatives for your commute based on current traffic
reports and data. These are just a few of millions of potential applica-
tions for artificial intelligence to improve our lives in a variety of
ways. A key part of this is in the data the systems have. If they have
enough, they can get smart enough to make recommendations that
are reliable and valid from a scientific perspective.
34 ARTIFICIAL INTELLIGENCE FOR HR

One way data scientists and developers are able to gather enough
data to drive artificial intelligence algorithms is by going beyond
explicit data to inferred data, opening up a vast amount of informa-
tion the algorithm can consume. As an example, let’s say you ask
someone about their favourite type of confectionery and they tell you
what it is. That’s explicit data, because it was clearly asked and
answered. However, if you scan someone’s social profile and see
dozens of photos of the person with a dog, you can infer that they are
a pet lover. While they may not have explicitly told you, it could be
deduced by looking at the evidence and coming to a rational conclu-
sion. Not to go too deep, but it’s worth pointing out that explicit data
is often more heavily weighted than inferred data, because it is a
truth, not an educated guess.
Algorithms can make both kinds of determinations using data. A
system can leverage things you tell it specifically, but in some instances
it can also examine data that may indicate something without you
having to say it outright. If an algorithm can consider those inferred
types of data, it can begin making broader decisions using hundreds
or thousands of data points instead of just the ones we select outright.
For computers to get better at making predictions and learn from the
data over time, they require immense amounts of data in order to
accurately process information. This isn’t about looking through one
hundred or even one thousand data points to train an algorithm. It
takes exponentially more.
Consider the example of the 2012 experiment at Google where
engineer Andrew Ng worked to teach a computer to understand
whether an image included a cat or not. Consider that for a moment.
This would be an easy task for virtually any child to complete –
simply look at the photo and determine whether a cat is present in
the photo or not. While it seems like a simple premise, the setup
required images from more than ten million videos in order to
adequately and accurately train the system to recognize cats within
photos with any measure of accuracy.13
While it’s easier to mention AI in marketing hype than it is to lever-
age it in practice, there are some technology providers that stand out
from the pack with rigorous approaches, complex algorithms and a
 THE BASICS OF ARTIFICIAL INTELLIGENCE 35

strong focus on improving the lives of HR professionals with intui-

tive technology. There is no shortage of software companies that are
pointing to their use of artificial intelligence, but how many of them
are really solving problems with the tools they’re developing? And
more importantly, how many of them are introducing bias into the
world either knowingly or unknowingly? Bias is a major concern
with AI, and this concept is explored deeply in Chapter 8 to help you
understand how bias can be programmed into an algorithm and our
responsibilities as business leaders for tackling that problem.
The following definitions are provided as a tool to help readers
ensure that they grasp some of the more nuanced applications and uses
for various aspects of artificial intelligence. They provide a foundation
of understanding for the discussion around specific applications of AI
in the workplace, from matching candidates to jobs to predicting what
kinds of training might be best for an employee’s development plan.

Machine learning
Machine learning is a type of AI that provides computers with the
ability to learn without being explicitly programmed. The problem
with computers historically is that they can complete a task, but then
their ‘mind’ resets and you have to start from scratch if you need
them to repeat that task. Machine learning works by examining large
sets of data and uses patterns in that data to improve a program’s
own understanding. In traditional machine learning, the learning
process is supervised and the programmer has to be very specific
when telling the computer what types of things it should be looking
for when deciding based on past known outcomes. For instance, the
programmer might explicitly tell the algorithm where to find new
data inputs, how to compute them and what to take action on. In
advanced machine learning, unsupervised algorithms learn from
inference and not from the programmer. This is when things truly get
interesting, because computers might see themes or other important
factors that humans simply can’t comprehend, and in supervised
machine learning the programmer might not take those factors into
consideration. With an unsupervised algorithm, the system can return
36 ARTIFICIAL INTELLIGENCE FOR HR

all relevant, interesting data points for consideration before adjusting

itself. While this is a neat idea, it’s not yet a reality for the most part.
Today virtually all commercially viable machine learning applica-
tions use supervised learning. As mentioned previously, the learning
element is the critical component. Many of the technologies laid out
in this book depend heavily on this aspect of artificial intelligence as
a foundational element of success. When machine learning truly
becomes unsupervised, it creates opportunities for greater learning
and understanding to occur.
Many large businesses are making headway in leveraging machine
learning. In a 2015 meeting, Google CEO Sundar Pichai said,
‘Machine learning is a core, transformative way by which we’re
rethinking everything we’re doing.’14 For a firm like Google, this
means every data point the company captures across its products is
used to guide and develop new capabilities and enhance others. For
instance, the more than 2 trillion annual searches in Google’s world-
renowned search engine, clicking activities across the company’s
global advertising network, and even how users interact with Gmail
all factor into the algorithms that seek to understand the human
population. The same goes for other Google-level technology compet-
itors such as Facebook, IBM, Microsoft and others. Each of the firms,
in its own way, is looking for ways to capitalize on the advancements
machine learning offers.
Clearly, machine learning has a multitude of applications and
many more that are yet to be defined. Some highly specialized forms
of machine learning are available today in the consumer context. For
instance, your video streaming service might suggest movies that you
will enjoy based on previous viewing patterns and ratings. Or your
favourite ecommerce platform might be able to predict what prod-
ucts you might like to buy based on a combination of your past
purchases, items you’ve explored in a search engine, and what others
with similar buying behaviours typically seek. These are common-
place and actually anticipated by consumers, and each relies on a
large amount of data to ensure their operation and accuracy of
predictions.
 THE BASICS OF ARTIFICIAL INTELLIGENCE 37

Natural language processing

Natural language processing (NLP) is a fundamental element of arti-
ficial intelligence. For machines to interact with humans, they must
understand how humans interact with other humans. In virtually
every case, that interaction is either written or verbal. Natural language
processing is the ability for a computer to understand human language
as it is spoken – the ‘natural’ part is essential. In order for a system to
understand human speech or text, it requires large inputs to train and
teach the system what humans sound like. Natural language process-
ing algorithms are based on machine learning, where interpretation
can be made from data that can be ambiguous, like spoken languages
typically are. Additional elements other than the specific words, such
as context, tone and structures in the data, can be interpreted by the
computer during the process. Instead of hand-coding large sets of
rules for how to read and understand speech, natural language
processing can rely on machine learning to automatically learn these
rules by analysing large sets of examples.
For a practical example, the OpenAI organization has made some
headlines in recent years with its developments of NLP technologies.
In one instance, the researchers told the algorithm to generate a fake
news story.15 With a few tries, the system was able to generate a fairly
believable story about scientists discovering living unicorns in a
hidden valley in South America. The system, called GPT-2, used 1.5
billion parameters to generate new text. The next iteration, GPT-3,
uses 175 billion. Not only will it get smarter and more capable over
time, but it will be able to create text faster than humans can. Scientists
are concerned about its ability to create stories, such as the unicorn
example, that are fake and that may mislead the public if they are
written at the same level as a human author.
NLP is commonly used today for a variety of tasks in our everyday
needs. One element of NLP that has promising applications is not
just the processing of the inputs but the recommendation of outputs
through automated question answering. For instance, the newest
version of Android’s stock Messenger application can read a text
from someone and suggest a handful of potential responses in order
38 ARTIFICIAL INTELLIGENCE FOR HR

to speed up the conversation. If someone sends you a text reading

‘How about dinner tonight?’, the application may offer up options
such as ‘Sounds good’, ‘Yes’ or ‘Maybe later’. Because NLP consumes
large amounts of data, it can already begin predicting potential
responses to questions with some measure of accuracy.
Some of the other use cases for natural language processing include
text mining, speech tagging and parsing, translation, sentiment analy-
sis and speech recognition. Text mining is a process of analysing text
data for trends and other insights. For employers, this might mean
seeing a list of keywords or phrases in your Glassdoor profile that illu-
minate how candidates and employees see your company and its
culture based on open text reviews, comments and ratings. Translation
is another obvious value point. In the past translation required a paid
translator and additional time for translation. While the output isn’t
100 per cent accurate at this time, online tools can translate many
languages instantly as you type. One final example to explore is senti-
ment analysis. Sentiment analysis is the process of looking at large
chunks of unstructured data, such as a sampling of correspondence
from a company’s email server, and determining the overall sentiment
or mood of the population based on that sampling. The process works
by analysing word use and tone, and extrapolating to determine the
perceived mood of the senders. We will explore this concept more fully
in the next chapter as there are tools available today to offer this value
to employers – the main purpose here is to show the value natural
language processing can bring to the workplace.

Deep learning
Deep learning is the next step, and possibly the most critical aspect on
the AI continuum; it is where machines begin creating true artificial
intelligence. At its most basic level, deep learning can be thought of as
a series of machine learning decisions where outputs from one deci-
sion inform the analysis of the next. At its core, human learning is a
set of processes that continuously sort through complex abstractions
by building a hierarchy in which each sequential level is created with
knowledge that was gained from the preceding layer of the ­hierarchy.
 THE BASICS OF ARTIFICIAL INTELLIGENCE 39

For instance, if you are shown a photo of an animal, you start filtering
it automatically and subconsciously without even thinking:
●●
Does it have fur, scales or feathers? Fur.
●●
What colour is it? Black.
●●
What is it? Cat.

The same process happens on a computational level in deep learning

where algorithms are stacked in a hierarchy of increasing complexity
and abstraction, mimicking the human learning processes. Computer
programs that use deep learning go through much of the same learn-
ing process. Iterations continue until the output reaches an acceptable
level of accuracy. The number of data processing layers necessary for
the process is what inspired the term ‘deep’ in deep learning. Not only
is deep learning fast, it is usually highly accurate. In order to achieve
a higher level of accuracy than machine learning, deep learning
programs require access to immense amounts of training data and
processing power, neither of which were easily available to program-
mers until the era of big data and cloud computing. Additionally,
deep learning programming is able to compare, analyse, create and
comprehend unstructured data.
Use cases today for deep learning include many types of big data
analytics like language translation, medical diagnosis, stock market
trading signals, network security and image identification. An exam-
ple of this in action is image identification, or computer vision. This
is the process of teaching a computer to understand what is in an
image. The experiment by Andrew Ng mentioned earlier in this
­chapter is a great example of how this works in practice. By teaching
a system how to ‘see’ a cat within an image, it became increasingly
good at discerning whether a cat was present in an image, regardless
of whether it was being held by a person, was partially obscured by
an object, or had some other distractor element in the image. Cats are
one thing, but it’s worth noting that this underlying technology has a
wide variety of applications. For instance, as companies continue to
work on the development of self-driving cars, one critical aspect will
be the identification of street signs and signals, and both vehicular
40 ARTIFICIAL INTELLIGENCE FOR HR

and foot traffic. Just like with human drivers, self-driving cars will
have mere fractions of a second to calculate whether a red light is
showing or a pedestrian has stepped into the road.

Neural networks
To scale deep learning, neural networks are the next step in the
complexity of artificial intelligence. A neural network is a system of
hardware and/or software loosely patterned after the operation of
neurons in the human brain. Within a human brain, there can be
billions of neurons that guide behaviours, impact learning capability
and more. Artificial neural networks are a variety of deep learning
technologies and operate by filtering decisions through a matrix of
layered calculations before arriving at an answer. Importantly, these
systems contain a learning element where they improve outputs over
time. The most common is the delta rule, where delta stands for the
range between the correct response and the successive responses prior
to it. In plain terms, a neural network is presented with a problem
and it makes a simple guess as to the answer, then it factors in the
delta and adjusts its algorithms with each following calculation to get
closer approximations to the correct answer. These calculations
happen rapidly, and once the neural network has been trained it can
serve as an analytical tool for other data.
Commercial applications of these technologies thus far include
handwriting recognition for cheque processing, speech-to-text tran-
scription, weather prediction and facial recognition. Facial recognition
is the one that most of us are familiar with, as both Google Photos
and Facebook have the capability to tell who is in a given set of
photos. If you upload a picture to your Facebook profile, it can
usually automatically tag your friends in the photo because it has
mapped their faces and recognizes them in the image. I’ll repeat my
earlier statement: while this sounds like a relatively simple task for a
human, it requires intense power and amounts of data for computers
to learn to accomplish this task. For example, what if only half the
face is on the screen? What if the person is wearing a hat? What if the
person has a very close relative that looks similar? In order to
 THE BASICS OF ARTIFICIAL INTELLIGENCE 41

­ etermine who a person is in a given photo, the algorithm must break

d
down the face into its component pixels, the tiniest parts of an image,
and then find similar pixel maps to compare and contrast until it
finds the right person. This might be easy with a pool of fifty or a
hundred people, but remember that Facebook has more than two
billion active monthly users it’s always crunching data on. Impressive.

HOW DO VENDORS AND EMPLOYERS KNOW WHAT AREAS

TO TARGET WITH AI AUTOMATION?

There are dozens and dozens of specific processes and practices that exist
within the HR function, and that can increase dramatically for enterprise
firms, those with unique compliance requirements, or even those operating
in multiple countries. So, which of those processes are best suited to being
automated? Which ones are prioritized? The answer is a logical one: focus
on high-volume, highly manual processes that have a high occurrence of
mistakes. This set of requirements helps to clarify what employers can
expect to see as machine learning algorithms continue to be leveraged for
process automation.
In a research interview with Aarti Borkar, Head of Offering Management
and Design for IBM’s Kenexa and Watson Talent products, she explained
that the company’s approach to artificial intelligence applications focused
not just on highly manual processes but also on those that were specifically
error-prone when operated by a human. This was a lightning bolt moment as
it helped me to understand how AI will progressively impact HR as a
profession. It’s not just that certain practices are more manual than
others – that is a given. It’s that some practices are more likely to have
decision errors, mistakes or omissions due to data velocity, volume and
variety. Some examples of those processes high on the list:
●
hiring and selection
●
workforce scheduling
●
payroll

Over time, AI applications will make their way deeper into HR process
territory, but for now, several of these highlighted areas are where
AI-developing firms will set their sights for the highest immediate impact
and scalability.
42 ARTIFICIAL INTELLIGENCE FOR HR

Types of artificial intelligence technologies

Because we do not have general purpose AI in place today, each piece
of technology serves a specific purpose, such as competing in a game
or responding to chat interactions. The use cases are many and varied,
but it often helps to explore the various examples in order to grasp
the applications and opportunities presented by artificial intelligence.
More than a dozen examples have been mentioned in this chapter,
some of which might have surprised you when you learned they were
powered by some form of AI technology. According to a report by
PitchBook, a venture capital investment analyst firm, venture capital
investments in artificial intelligence technologies have exploded in
recent years. The trend is up twelve times from what it was in 2008,
with $6 billion invested across more than 600 deals in 2017.16

TABLE 2.2 A sampling of artificial intelligence technologies

Example Type Result Technology

AstuteBot Chatbot Customer service bot answers Natural language
questions processing, machine
learning

Alexa, Voice-based Ask questions, get answers; over Natural language

Cortana, Siri assistants time, answers become more processing, speech
personalized based on activity processing, machine
and signals learning

Microsoft Learn and imitate System chats with people after Natural language
Tay how humans they begin interacting with it processing and
interact online machine learning

Alibaba Reading Read questions, predict answers Neural network,

comprehension on Stanford University reading/ natural language
comprehension test processing

Deep Dream Replicates human Learn as a human learns. Neural network

brain processing Current project is finding and
categorizing all internet imagery

Tesla Self-driving Car that is aware of Machine learning

Autopilot vehicle surroundings and can adjust
course to avoid accidents
 THE BASICS OF ARTIFICIAL INTELLIGENCE 43

Table 2.2 offers an illustration of the range of complexity that appears

in artificial intelligence. The spectrum moves from broadly applicable
examples to highly specialized systems, some of which you might be
familiar with in your everyday life.

AI on a global scale
There is a tremendous amount of support for artificial intelligence
around the world, as it promises new applications and improvements
in existing tools. Self-driving cars would rely heavily on AI, as would
automated tools that screen MRIs for abnormalities. Each country
has a different culture and set of values, which means each will tackle
the problem of developing rich AI capabilities with a different mind-
set and fervour. Within the context of this book we talk about many
startups and other human capital management (HCM) technology
companies, many based in the United States simply because the
employers that demand advanced human capital technologies are
often also based there, offering a target-rich environment for vendors
and service providers. However, investment in and further develop-
ment of AI is a global phenomenon. In a research interview with
Mahe Bayireddi, CEO and co-founder of Phenom, he explained his
perspective on this as the leader of a company serving talent and HR
leaders in dozens of countries:

The global appetite for AI-enabled HR tech is universal, but its

application will always depend on context. First and foremost, we
should be sensitive to different countries’ regulations and cultures. AI
is not a one-size-fits-all solution, and its growth in any market will be
a mutually educational process for both the provider and beneficiary.
We’re seeing enterprises pay close attention to the localization of AI,
which means its advantages are rapidly crystallizing around the world.

Bayireddi’s company, Phenom, is a talent experience platform with

embedded AI capabilities, including chatbots, smart notifications,
and other tools to augment recruiters and automate parts of the
hiring process. The company is based in the United States but has
44 ARTIFICIAL INTELLIGENCE FOR HR

expanded to serve customers in more than 130 countries over time.

As Bayireddi points out, the needs are universal but there will be
local shaping and context depending on culture and other aspects.
Below some examples are shared about where AI development is
taking place around the world.

China
China is one country that has publicly declared its intent to invest
heavily in AI. The country’s plan calls for the nation to be the world
leader in AI by 2030, which has spurred heavy investments in AI start-
ups and infrastructure. It’s entirely possible that China may in fact
reach that goal, because the government allows some activities that
other countries might not. For instance, public video footage from
police cameras in Guangzhou is being provided to facial recognition
startups to support data analysis and algorithm development.17 We
know that the fundamental aspect of being able to create AI tools that
are usable and useful is a ready, and large, source of data, and the
Chinese government is willing and able to provide that data free of
charge. Another step in the pursuit of global AI domination came in an
announcement in early 2018. Chinese company Alibaba made head-
lines when it announced that its AI system was able to beat a human
score for the first time in the Stanford University reading and compre-
hension exam, a test developed by Stanford scientists and based on
data available in Wikipedia.18 The system’s neural network was
required to crunch the data and respond to more than 100,000 ques-
tions. To be fair, Microsoft’s own algorithm also beat the human score
in this specific competition, but the results took an additional day to
compute and finalize, meaning China took the gold in this event.

The United Kingdom

The UK is no stranger to AI technologies either. In 2017 many new
startups were highlighted in the media as they launched in order to
tackle any number of problems, from self-driving vehicles to cyber
security to background checks. Here are brief examinations of just
 THE BASICS OF ARTIFICIAL INTELLIGENCE 45

three of the systems to demonstrate the wide variety of use cases and
opportunities to leverage AI in today’s fast-paced world of technology
development.19 For starters, Onfido is a system designed to use facial
recognition technology to support identity verification and background
screening. Historically this practice has been accomplished by collect-
ing sensitive personal data and can take several days to complete the
process, potentially exposing users to identity theft as their data is
provided, gathered and submitted. Another tool, BenevolentAI, uses
deep learning to mine data from successful drug creation and discovery
practices to shorten the time to market for new, potentially lifesaving
medications. One of the key factors in the high price of medications is
the significant lead time and investment in discovery and testing –
shortening that period could potentially reduce costs for new
medications, making them more accessible to those in need. Finally,
OakNorth is an application designed to measure risk. While that
would be valuable in any number of industries, it is most beneficial to
the financial industry. Leveraging AI, the system would identify and
price risks in such a way that firms could make faster, better decisions
about credit and lending. In the future, this type of system could
augment or even replace actuaries, the individuals that calculate risks
and influence prices and offerings in the industry.
Other countries around the world are also racing towards the devel-
opment and adoption of AI-based technologies, including France,
Israel and Canada, to name a few. The takeaway here is that AI tech-
nologies are here to stay. The use cases are broad, deep and varied,
and there is no way to even begin to predict how much of an impact
these smart technologies will have on the way we live and work. The
good news is we don’t have to know how all of the technology is
developed. We don’t have to build our own algorithms and systems.
We just need to be really clear about the problems we have and the
kinds of solutions that can improve the experiences for candidates,
employees and business leaders. While the rest of this book might
delve into technical bits here or there, for the most part we are done
talking about the nuts and bolts of AI technology. Now, let’s dive into
how AI can apply to the world of HCM.
46 ARTIFICIAL INTELLIGENCE FOR HR

ADDITIONAL RESOURCES

As promised at the beginning of this chapter, if you are technically minded

or interested in diving deeper into the technology components of artificial
intelligence, I would recommend exploring the following resources
alongside the exhaustive list of reference material I’ve cited throughout the
book. Each of the resources below is full of insightful information from a
perspective outside HR, which means you’ll spend some energy thinking
about how these translate from a broader point of view to the one that
matters most to you. Experience has taught me that’s a valuable exercise to
participate in.
●●
Stanford University’s One Hundred Year Study on Artificial Intelligence
(AI100) – Analyses and examines the potential impacts of AI on the
world, including life, work and entertainment. https://ai100.stanford.
edu/
●●
Prediction Machines: The simple economics of artificial intelligence – This
book examines the economical side of AI, including how the cheaper
cost of predictions has led to more applications and examples of AI
being used.20

KEY POINTS
●●
AI is the next logical step in business evolution now that we have the
capability to capture, store and analyse more data than ever before.
●●
AI technologies include machine learning, natural language processing
and deep learning. Each has use cases that affect our lives in both
consumer and workplace contexts.
●●
Adoption and investment in these technologies is rapidly increasing, and
governments and employers around the world are aggressively pursuing
research and development of advanced systems powered by AI.
 THE BASICS OF ARTIFICIAL INTELLIGENCE 47

Notes
1 J Copeland. Alan Turing: The codebreaker who saved ‘millions of lives’, BBC,
19 June 2012. www.bbc.com/news/technology-18419691 (archived at https://
perma.cc/UR4X-B558)
2 N Sharkey. Alan Turing: The experiment that shaped artificial intelligence,
BBC, 21 June 2012. www.bbc.com /news/technology-18475646 (archived at
https://perma.cc/XQ8T-2TTK)
3 S Harris and E Spenser. Sierra Cedar 2018–2019 HR Systems Survey: 21st
Annual Edition, 2018. www.sierra-cedar.com/wp-content/uploads/Sierra-
Cedar_2018-2019_HRSystemsSurvey_WhitePaper.pdf (archived at https://
perma.cc/99V6-KYMM)
4 M Edwards and K Edwards (2016) Predictive HR Analytics: Mastering the
HR metric, Kogan Page, London
5 Lexico. Artificial intelligence, 2021. https://www.lexico.com/definition/
artificial_intelligence (archived at https://perma.cc/8ME3-7XAM)
6 P Stone, R Brooks, E Brynjolfsson, et al. Artificial Intelligence and Life in
2030. One Hundred Year Study on Artificial Intelligence: Report of the
2015–2016 Study Panel, 2016. https://ai100.stanford.edu/sites/g/files/
sbiybj9861/f/ai100report10032016fnl_singles.pdf (archived at https://perma.
cc/9WUZ-A9PR)
7 A Agrawal, J Gans and A Goldfarb (2018) Prediction Machines: The simple
economics of artificial intelligence, Harvard Business Review Press, Boston, MA
8 D Epstein (2019) Range: How generalists triumph in a specialized world,
Macmillan, New York
9 V S Rotenberg. Moravec’s Paradox: Consideration in the context of two brain
hemisphere functions, Activitas Nervosa Superior, 2013. 55, 108–11. https://
link.springer.com/article/10.1007/BF03379600 (archived at https://perma.cc/
TAN5-YZFV)
10 K Bailey. Reframing the ‘AI effect’, Medium, 27 October 2016. https://medium.
com/@katherinebailey/reframing-the-ai-effect-c445f87ea98b (archived at
https://perma.cc/3H9B-KR3C)
11 L Tesler. CV: Adages and coinages, n.d. www.nomodes.com/Larry_Tesler_
Consulting/Adages_and_Coinages.html (archived at https://perma.cc/
JX44-ZZF2)
12 M Marking. How big tech is using machine translation as an AI proxy, Slator,
7 October 2016. https://slator.com/technology/big-tech-using-machine-
translation-ai-proxy/ (archived at https://perma.cc/L3VU-7QHD)
13 J Markoff. How many computers to identify a cat? 16,000, The New York
Times, 25 October 2012. www.nytimes.com/2012/06/26/technology/in-a-big-
network-of-computers-evidence-of-machine-learning.html (archived at https://
perma.cc/C9R2-8BQG)
48 ARTIFICIAL INTELLIGENCE FOR HR

14 J Niccolai. Google says it’s ‘rethinking everything’ around machine learning,

PC World, 23 October 2015. www.pcworld.com/article/2996620/business/
google-reports-strong-profit-says-its-rethinking-everything-around-machine-
learning.html (archived at https://perma.cc/HT6V-J3EY)
15 EC Baig. Too scary? Elon Musk’s OpenAI company won’t release tech that can
generate fake news, USA Today Tech, 15 February 2019. www.usatoday.com/
story/tech/2019/02/15/elon-musks-openai-fake-news-generator-too-scary-
release/2880790002/ (archived at https://perma.cc/R749-XPEP)
16 C Stanfill. Real potential for AI, PitchBook, 2018. http://files.pitchbook.com/
website/files/pdf/PitchBook_1Q_2018_Analyst_Note_Real_Potential_for_AI.
pdf (archived at https://perma.cc/5PHR-PVXT)
17 J Horwitz. The billion-dollar, Alibaba-backed AI company that’s quietly
watching people in China, Quartz, 16 April 2018. https://qz.com/1248493/
sensetime-the-billion-dollar-alibaba-backed-ai-company-thats-quietly-
watching-everyone-in-china/ (archived at https://perma.cc/JFC9-YMSG)
18 Lucas. Alibaba and Microsoft AI beat humans in Stanford reading test, The
Financial Times, 2018. //www.ft.com/content/8763219a-f9bc-11e7-9b32-
d7d59aace167 (archived at https://perma.cc/L882-3NCE)
19 Nanalyze. Top-10 British artificial intelligence startups in the UK, 4 May 2018.
www.nanalyze.com/2018/05/top-10-british-artificial-intelligence-startups-uk/
(archived at https://perma.cc/LWP3-Z6MH)
20 A Agrawal, J Gans and A Goldfarb (2018) Prediction Machines: The simple
economics of artificial intelligence, Harvard Business Review Press, Boston, MA
 49

03

General AI applications within HCM

So you’ve finished your employee satisfaction/engagement survey and

you’re ready to report on the results. What’s the average engagement?
In the 1940s, the US Air Force was losing planes faster than it
thought possible. It wasn’t because of any wartime activity or identi-
fiable technical issue, which was why the problem was so puzzling to
leadership. It formed a team to measure the pilots, wondering if the
size of 1940s’ aviators was different from those standard sizes for
which the planes had been designed years before. The results of the
measurement of over 4,000 pilots on 140 different dimensions led to
a depiction of the ‘average’ pilot that was to be used for future design.
However, one of the researchers wanted to go deeper before accept-
ing this idea of an ‘average’ pilot size. Lieutenant Gilbert Daniels
proceeded to examine the data for the ten dimensions most relevant
to cockpit size, including measurements like height, sleeve length and
chest circumference. He then expanded the definition of ‘average’ to
the middle 30 per cent of each measurement to make it more inclu-
sive. To his surprise, he found that not a single one of the 4,063 pilots
measured within the average range for all ten dimensions. None.
Zero. In practical terms, this completely broke the idea that there was
some ‘average’ size that described all, or even a majority, of pilots.
In the world of business, average is a mathematical term, not a
human term. And yet in many organizations today we carry out
engagement surveys and other activities, reporting back on the aver-
age this or the average that, never realizing that we’re likely quite far
off the mark when it comes to how any particular individual might
50 ARTIFICIAL INTELLIGENCE FOR HR

be feeling. That’s why I was so intrigued by the opportunity to run an

experiment allowing humans to test their mettle against an algorithm
to see which was better at understanding employee survey data.
Spoiler alert: AI is better than humans at understanding employee
surveys.
That’s the big takeaway from one of our team’s research experi-
ments performed in 2019.1 For the study, we gathered 1,000 responses
from North American workers about their work relationships, chal-
lenges, stressors and more. We not only asked about how they felt on
a scale (very satisfied, satisfied, neutral, dissatisfied, very dissatisfied),
but we also asked open-ended questions and gathered a tremendous
amount of free text to analyse. Then came the hard part.
We passed the data to some amazing consulting professionals that
spend their time analysing employee survey data. We told them that
we ran a 1,000-person organization and needed to know what issues
were present within the firm that needed addressing. Then we took
that same data set and ran it through an algorithm that examines
survey data for words that are associated with employee sentiment
and emotions. The responses were astonishing. In the consulting
responses, we got some broad insights and overarching themes (e.g.
70 per cent of your employees are stressed at work), and it was gener-
ally helpful. However, only one of the consultants even tried to look
at the open-text responses in the survey, and only at a cursory level.
That’s what made the algorithm’s performance so incredible by
comparison. The AI system was able to shed light on things at a much
deeper and more personalized level. For instance, it showed differ-
ences such as:
●●
the difference in stress levels for those who were white versus those
who were people of colour
●●
the gap between how men and women felt about their bosses
●●
the variation between employees who worked in one job or
location compared to another

Not only did we get more specific insights from the AI tool, but we
also were able to see the emotions and mood of the employees in
 GENERAL AI APPLICATIONS WITHIN HCM 51

their responses, giving even further detail into how they perceived
their boss and their work. What excited me most about this was
thinking about how we as HR leaders can finally act on these kinds
of open responses from employees instead of ignoring them or only
getting a very basic surface level view of things. Additionally, we
found that some of the findings from the human analysis could be
biased by existing notions. For example, if someone reviewing quali-
tative comments on an engagement survey has a difficult boss, then
every response referencing a manager will dredge up feelings of
discomfort and reinforce the person’s belief that managers are a
problem. The AI didn’t have any preconceived notions like that in
this experiment.
Less than two months after finishing this experiment, I spoke with
an HR executive for a Canadian firm with tens of thousands of global
workers. He told me confidentially that the firm always asks open-
ended questions on employee surveys but that the team doesn’t have
time to read them and hasn’t acted on them in years. For extreme
cases like that and for everyday uses, AI is better than humans at
understanding engagement surveys. If that gives you pause, don’t
worry. AI can’t solve the problems. That’s where we come in.
In this chapter we’ll look at how AI can be used in some broad
ways within HR, how its popularity is catching on within the indus-
try, and more. This is a great transition from AI’s definition in the last
chapter and the rest of this book, where we will examine deep, prac-
tical use cases for algorithms in our everyday work.

AI solves administrative HR problems

One of my favourite quotes from an old friend is that great HR is
invisible. If we’re doing our jobs correctly, then employees are enjoy-
ing their work, getting along with their managers and performing at
their most productive levels. I’d argue that a business has never
succeeded wildly and then pointed at consistent and effective HR
practices as the key to success, even if that’s true. We’re focusing
instead on the actual results, not the enablers that made it possible.
52 ARTIFICIAL INTELLIGENCE FOR HR

Similarly, one of the concepts I’ve found in my research is that when

AI works correctly, it’s virtually invisible, because we’re so focused on
the end result (not the technology that made it possible). When AI
works as it’s supposed to, it can be relatively boring. The results might
be exciting, but the AI component itself is of little interest once the
novelty wears off. That’s not necessarily a bad thing, because the
priority for us as employers shouldn’t be the technology – it should be
the outcome. Take self-driving cars, for example. If you’ve ever ridden
in one, the experience is a little bit, well, boring. After the novelty
wears off, you lose interest just as you would if you were riding in a
taxi or other vehicle where someone else is handling the driving. The
reason this is a challenge is that AI might be seen by some as a fad. The
very concept of AI might be seen as just another ‘flash in the pan’
instead of the revolutionary technology that it is. I’ve been very care-
ful in this book to talk about AI not for its flashy ‘fun’ value but in the
context of the very real problems it solves for you, for me and for
businesses around the world. Within this chapter, the goal is to intro-
duce you to some of the ground-breaking technologies available today
as well as some of the ways they may affect how HR is handled for
years to come.
Regardless of the value of the technology, today HR is a highly
administrative function. Yes, we want to and often do impact the
organization in strategic ways, but we still have to contend with
employee files, paperwork and compliance tasks that require our
attention. We explored some of the various technological underpin-
nings in the last chapter and will look at some of the important use
cases spanning the full employee life cycle in forthcoming chapters,
but this specific portion of the book is geared towards the adminis-
trative components and general applications for AI technologies in
the workplace. Some of these might bleed over into other administra-
tion-heavy types of functions like legal or accounting, but because
they are an essential part of HR I didn’t want to skip over them in my
rush to look at recruiting, learning or talent management in subse-
quent chapters. Don’t just take my word for it. In an interview with
 GENERAL AI APPLICATIONS WITHIN HCM 53

the Chief Data Scientist of Jobvite, Morgan Llewellyn, I was surprised

to hear this from a data-focused executive:

HR shouldn’t have to be AI experts, but HR should have a clear picture

of their strategic vision. I would encourage employers to take a results-
first approach as opposed to a technology-first approach. Have a clear
vision of the result you want to achieve and what success looks like for
your organization.

In the last chapter you saw how much money is flowing into AI tech-
nologies from an investment perspective: billions of dollars. And our
very own industry is getting its due as well. In our research over the
last four years, we have seen billions of dollars in funding across
nearly a thousand discrete funding events within the HR technology
space. These investments have spanned various rounds (seed, A, B, C
and beyond), technology focus areas (recruiting, learning, benefits,
talent and core HR) and countries (this is a global phenomenon).
In an in-depth analysis of the firms receiving funding, my firm
found that nearly half of the 2017 and 2018 investments in HR tech-
nology went to firms touting bots, machine learning and other AI
features and components.2 Between 2017 and 2018, there was an
increase of four times the amount of funding flowing into AI-focused
companies in the HR technology space. The AI focus in 2019 was
slightly lower as a percentage, but the overall amount of funding
dollars within the industry was even higher. While 2020’s numbers
were lower overall due to the global pandemic, the early numbers
from 2021 look like they will dwarf any of the previous years on
record. Bottom line: these investments in machine learning and AI
will lead to smarter and more capable technologies than ever before.
Based on the messaging I’m seeing, I expect this volume to increase
into the future, especially in critical automation areas such as sourc-
ing and screening candidates, highlighting employee sentiment, and
other highly administrative activities. Providers on the list of those
receiving funding ran the gamut from those focused on data analytics
or learning to those exploring hiring or employee engagement.
54 ARTIFICIAL INTELLIGENCE FOR HR

FIGURE 3.1 AI-focused HR technology firms receiving 2021 venture capital funding

Talent Learning &

acquisition development

Mya systems Sana Labs

Conversational recruiting Learning content

Celential StriVR
Automated sourcing Virtual reality

Performance Core HR &

& engagement workforce

Syndio Immedis
Pay equity Global payroll

InsideBoard Payactiv
Change management On-demand pay

SOURCE Lighthouse Research & Advisory, 2021

Figure 3.1 depicts a handful of the firms that were categorized as

having artificial intelligence technologies in their offerings. To provide
some context, these are some examples of firms that received funding:
●●
Phenom received over $130 million in funding during 2020 for its
talent experience management product. The system supports talent
acquisition and internal mobility use cases, helping employers
through the use of AI-driven tools that target candidates and
engage with them using chatbots and automation.
●●
Immedis, a global payroll technology provider based in Ireland,
raised $50 million in late 2020. The technology aims to use
automation and machine learning to solve the age-old problem of
paying employees in multiple countries, allowing employers to
focus on the more pressing tasks of managing and supporting a
global workforce.
●●
Gloat, an internal talent marketplace provider serving Unilever,
Deloitte, Schneider Electric and other large organizations, received
$25 million in one funding round. The company’s algorithms can
help to illuminate career paths for workers at a highly personalized
level, helping to drive engagement and retention among the workforce.
 GENERAL AI APPLICATIONS WITHIN HCM 55

As new funding rounds continue to occur, I expect to see additional

money poured into firms that are leveraging AI technologies like
machine learning and natural language processing to help automate
tasks, speed up solution delivery and personalize service for workers.
There is incredible value in these types of technology to change the
workplace landscape for the better.

What about the impact on jobs?

Virtually every AI-related discussion I’ve had with executives and
technology providers in the last few years has revolved around a
central question: what will AI do to our jobs? For instance, a 2016
survey showed that about two-thirds of respondents expected robots
to do much of the work of humans in approximately fifty years;
however, four out of five of those same respondents said, ‘But not my
job.’3 Did you catch the mathematical paradox there? Sixty-seven per
cent of people think jobs will change but 80 per cent say it won’t
change their own job. It’s a classic example of overvaluing the skills
we have and undervaluing the skills of others in an attempt to save
face, and I’ve pointed at this research in numerous presentations to
help show that we can be incredibly irrational when faced with fears
of job loss or irrelevance. Don’t worry, I don’t follow that logic. Not
only am I a technological optimist, I’m also a pragmatist. Every time
technology changes to the degree it starts to impact jobs, new jobs
are created. It also changes the demand for skills, which I will explore
in detail in Chapter 9. While news items like this survey make for
interesting headlines, in reality there’s no way for the general public
to know how much of an impact robots, algorithms and AI will have
in the next five years, much less fifty years into the future. To take a
step back and look at the question objectively, this kind of techno-
logical advancement has happened literally thousands of times
throughout history, from the earliest plough to the invention of the
personal computer. New technologies are created that disrupt and
change how an industry operates, and the people that are displaced
must find new roles and opportunities for employment.
56 ARTIFICIAL INTELLIGENCE FOR HR

To provide a more recent example, consider the ubiquitous use of

search engines today. Around the world, millions of searches per day
are leading people to articles, videos, social sites and other online
resources. How do they find what they are looking for? Search
engines. In order to help websites improve rankings and be seen by
more people, companies employ specialists in search engine optimi-
zation (SEO). Think about it: even as recently as 1990, search engine
optimization did not exist. Nobody held a job as an SEO expert
within an agency, as a freelancer or as a consultant. Today, a quick
search of a popular global job board shows more than 3,000 job list-
ings for individuals with search engine optimization expertise, and a
search on LinkedIn shows more than 500,000 professionals with
some variation of ‘search engine optimization’ in their job title or
skills. The parallel is clear: even as mature industries are disrupted,
new industries and opportunities are born.

ETHICAL POLICIES FOR HR TECHNOLOGY PROVIDERS

In 2019, I could only find one provider in the HR technology space with a
clear, established code of ethics when it comes to the use of artificial
intelligence in the product. Today, I can name many more providers that
have taken steps to emphasize their approach to AI as a complement to
human effort, not a replacement for it.
In addition, another ethical component, discussed in deeper detail in
Chapter 8 of this book, looks at algorithm bias. While the overall
components vary depending on the use case (hiring may necessitate
different approaches than learning and training), the overall ethical
approach includes a few key elements:
●●
Human-centric: People are in charge, not the algorithm.
●●
Transparent: The way the algorithm works is clear and explainable.
●●
Fair and equitable: The algorithm treats everyone equally, and every
effort is made to ensure no adverse impacts.
●●
Secure and private: The data is not only protected, it is private and not
repurposed.
 GENERAL AI APPLICATIONS WITHIN HCM 57

●●
Governance: The AI will be audited, evaluated and updated as needed to
ensure that the above items remain true and accurate.

For employers, this can help to put teams at ease regarding the intent to
replace workers with an algorithm. It also provides a common language and
set of expectations for what the algorithm should and should never do.

There’s definite optimism within the ranks of the HR profession

about the value artificial intelligence can bring. A 2017 study by the
Canadian HR Professionals Association found that 84 per cent of
HR professionals felt that AI was a valuable tool for supporting
human resources objectives, yet most HR leaders were not yet very
familiar with how AI operates and the type of value it can provide.4
To combat the vast amount of misinformation and hype in the
market, I’ve spoken in front of hundreds of corporate, education and
association audiences in the last few years about AI, how it works,
the core use cases for HR operations and how this will affect the
skills of the workforce.
Despite these positive beliefs, many of the initial conversations about
AI turn to how the adoption of these technologies will impact jobs
globally, not just within the HR profession. Whether for the shock
value or some other reason, it’s hard to enter a discussion that doesn’t
have this job loss component as one of the key storylines. However, for
the purposes of this conversation, we are going to focus more keenly
on the actual jobs of HR professionals in an attempt to understand the
degree to which this automation and/or replacement may happen.
Taking a more specific look at human resources, it’s easy to see ways
that the advance of AI could disrupt how businesses have operated for
quite some time. Human resources as a profession is relatively young,
but it is still heavily administrative in most companies.
From setting up employee taxes and benefits during the early days
of onboarding to sifting through hundreds of resumes a year for job
openings, the administrative, hands-on work of human resources is
ripe for disruption and automation. For instance, one recent client
58 ARTIFICIAL INTELLIGENCE FOR HR

my team supported during an HR technology selection was using

nearly 50 per cent of the HR team’s time to do duplicate work. Time
was tracked in a specific system, but the HR team then had to analyse
the time and fill any gaps by hand. Then, the team would manually
enter any vacation or sick leave into the system to ensure everyone
had the correct number of hours on their time sheet. After that, the
data was downloaded and then uploaded into the payroll processing
tool. That data from payroll, however, never made it back into the
main system of record. It stayed separate and siloed, and it could
never be reported on in a holistic fashion alongside any other perti-
nent HR data. This example shows ample opportunities for robotic
process automation to help. Not only would it increase efficiency, but
it would help to make the HR team more effective. Research shows
that most employers are planning to or already using some sort of
automation to support routine tasks.
But not everyone thinks that automation will change jobs, which
may leave them vulnerable to skills gaps and competition. In an
intriguing look inside the recruiting function, we performed a study
in late 2019 on talent acquisition leaders to understand their percep-
tions of how skills are evolving for today’s recruiting professionals. I
was quite puzzled to review the results and find that about half of
those responding said that newer skills like virtual collaboration or
analytical skills were not important for recruiters to develop. Upon
further inspection, I uncovered a fascinating connection. In another
question, we asked about how AI would impact recruiting. If a
respondent thought AI was mostly hype, they also said that the new
and evolving skills didn’t matter. If a respondent thought AI was
going to change recruiting and automate some of the more common
tasks, they also said that developing new skills was a key priority.
What we believe drives what we do, and for those that think AI is
more hype than reality, the future is going to be a rocky one indeed.
The concept of ‘swivel chair automation’, or robotic process auto-
mation, is an algorithm designed to pull data from one system and
insert it into another. This has a clear connection to the types of work
done in many HR organizations. Another tangible example of this
comes to us from a provider of automated chatbot tools for ­recruiting
 GENERAL AI APPLICATIONS WITHIN HCM 59

purposes. One client of the company was a large cosmetics firm look-
ing to hire a series of brand ambassadors with large social media
followings in order to capitalize on the reach and influence those
people would offer. In the actual chat interactions with candidates on
the website, the algorithm would ask the person about their most
active social media channel and username, and then it would auto-
matically go and pull in their data in the background, storing it with
their application. For example, if the person said that Twitter was
their preferred channel, the system could automatically check the
person’s Twitter handle and their number of followers and interac-
tions in recent history. Someone with ten thousand engaged followers
would be much more appealing as a candidate than someone with
twenty passive followers. This would be challenging and time-
consuming to handle as a physical human task, but the chatbot
algorithm accomplishes this virtually instantaneously.

Human interaction, non-human systems

If you were tasked with trying to help your workforce understand the
potential impacts of AI and robotic process automation on your
company, what would be your solution? For Guardian Life, an insur-
ance firm with 11,000 employees and service representatives around
the globe, the answer is to schedule a robot play date. Deanna
Mulligan is the Chief Executive Officer of Guardian Life, and she
knows that the world of insurance claims, actuarial tables and other
components are going to be some of the most easily automated
aspects of work. For employees this can be disconcerting, as they
don’t know what kind of impact this might have on their jobs.
According to Mulligan, she expects only about 10 per cent of the jobs
at the firm to be affected by technological advances, but she’s not
approaching this passively.5 The rest of the firm will be disrupted by
the technologies that are coming, and the best way to help the rest of
the workforce prepare for that future is to give them opportunities to
interact with and understand those technologies – hence the robot
play dates. These events allow back office workers to interact with
small robots and take robotics classes to learn more about how the
60 ARTIFICIAL INTELLIGENCE FOR HR

technologies work. More importantly, they see how the robots can
improve what they are doing and where the robots can’t take over
their roles of serving customers. The end goal is to help the workers
see that it isn’t about workers or AI coming out on top. Instead, it’s
about finding the best components of both resources and partnering
up to better serve the customer base.
I love this example because it shows that we don’t have to stick our
collective heads in the sand and hope that automation passes us by.
In fact, we can embrace it knowing that there are better outcomes
ahead. This applies whether you’re the CEO at a multinational firm
or you’re the head of HR for a company with 100 employees. There
is tremendous opportunity for automation to bring value into the
workplace, and HR is at the forefront of this trend and how it will
affect and impact our people. We are strategically positioned to
demonstrate for the rest of the business how we leverage these tech-
nologies and how they can improve the work environment for all.
One reality of adopting artificial intelligence is that workers, includ-
ing HR, will have to work alongside systems and processes that are
automated. How does this affect us? To what degree are we prepared
for this sort of work environment? I will admit that when I started
encountering bots for candidate conversations, I was a bit sceptical. I
mean, how valuable is that medium, really? It turns out that it creates a
fundamentally different kind of experience for people, and that’s a good
sign. For example, during a briefing with the team that runs Paradox.ai,
a bot designed to support recruiting and candidate engagement, one of
the company’s representatives unveiled a surprising statistic. Candi­
dates typically know they are chatting with a computer system called
‘Olivia’ during the application process, but at the end of the conversa-
tion, approximately 75 per cent of candidates actually thank the bot for
the interaction. It’s hard to know how much of this is simple routine,
but the fact remains that we will have to understand how the relation-
ship operates and to what degree we should humanize the algorithms
that interact with candidates and employees within these systems.
To make this point about how humans can interact with non-
human systems, let’s look at one of the biggest pain points for
 GENERAL AI APPLICATIONS WITHIN HCM 61

workers everywhere: scheduling meetings. One of the common tools

in use today by busy business professionals is a scheduling applica-
tion to manage their calendars and appointments. To avoid the
back-and-forth hassle created by attempting to schedule meetings via
seemingly endless volleys of email, these scheduling apps let users
synchronize calendars and then share that calendar with people they
want to meet with, making the process more passive and automated.
There are a variety of tools offering this functionality, and each brings
a slightly different approach to the audience and problem being
solved. Sales tools help to ensure that a customer database is updated
in sync with sales calls. Customer service tools can identify mutual
contacts between two individuals to increase the positive sentiment
between two individuals. Consultants can generate contracts based
from calls with pertinent customer data automatically added. These
elements don’t radically change meetings, but if they make them two
per cent or ten per cent better, why wouldn’t we use them? The way
this technology operates is simple: users leverage their favourite
messaging platform to interact with the system, submitting requests
that are then managed and handled by the algorithm. Messaging
platforms can be anything from Microsoft Teams and Slack to
Facebook Messenger, Skype or even SMS text messages.
In the end, for many workers it comes down to productivity. Can AI
help us to be more productive? One research study estimated the cost
of lost productivity due to non-job-related administrative work at
$4,600 per employee.6 Imagine for a second that you run a company
with 1,000 employees. On average, workers at your firm would be
wasting $4.6 million on those kinds of tasks that might not even add
that much value to the business. The more we can give these tasks to
AI, the more we can turn our employees towards things that impact the
business. Things that drive value for customers. Things that matter.

Types of broad-spectrum AI components in HCM

In the last chapter we covered some of the basics of how machine
learning and natural language processing work. Now we can begin
to analyse types of broad-spectrum AI components that can be
62 ARTIFICIAL INTELLIGENCE FOR HR

TABLE 3.1 Task prioritization matrix

Urgent Not urgent

Important Do Plan

Not important Delegate Eliminate

leveraged within the human capital management function. While

these continue to develop every day, at the time of writing there are
already dozens of systems focused on a variety of HR tasks being
automated and augmented with AI. Remember, the tasks to prior-
itize with automation are those that are high-volume and repetitive.
The more specialized a specific task might be, the better. Some tasks
are urgent – they need to be done quickly. Other tasks are impor-
tant – their completion has a critical impact. If you think about this
as a four-quadrant grid (Table 3.1) mapped to urgency and impor-
tance, this allows someone to complete more work than ever before
in those areas that are important but not urgent, or urgent but not
important.
Historically, an individual might have delegated a task that was
urgent but not important, such as scheduling meetings or gathering
employment contract data, to a person. Today, automated systems
can easily accomplish these and other similar tasks with a chatbot.
Within a few seconds the system can handle what previously took
anywhere from a few minutes up to half an hour, depending on how
systems are organized and documents are stored. Similarly, while
items in the lower right quadrant might have been completely ignored
or overlooked in the past, it’s possible that some of those actions are
now able to be handled by a computer. For instance, reading through
every piece of employee feedback and synthesizing a few key points
and themes from the content would not have been a task most
employers would be willing to pay a worker to do, especially if the
firm has thousands of workers. Yet systems such as the ones high-
lighted below can do this rapidly in near real-time. There are literally
hundreds of uses spanning almost any task that an HR leader might
have to accomplish, so let’s explore some of the more pertinent ones
that have advanced pretty quickly in terms of development.
 GENERAL AI APPLICATIONS WITHIN HCM 63

AUTOMATION CREATES TIME: AVOIDING PARKINSON’S LAW

Have you ever had a month to work on a project that you could finish in a
few hours, only to wait until the last minute to get it done? This common
human phenomenon is known as Parkinson’s Law, which says any task we
have will expand to fill the time we give it. In the context of artificial
intelligence, one of the points of value is new time efficiencies generated by
automation. Or, in the standard language everyone repeats, ‘AI will free up
more time for strategic work.’
Now I fundamentally believe that, but I also know that we’re dealing
with human nature. If we don’t plan for the time, the margin, that
automation can create, we’ll find ourselves filling our newly available time
with unimportant tasks or busy work. If you are looking at technology to
support your team, whether in hiring, talent development, core HR or
another area, it’s important to think about how your team will use that
time. In a research study early in 2021, we filtered our data for companies
that say they are seeing efficiency benefits from automation, AI and similar
tools. We then asked how those recruiting teams were spending that
additional time. The top answer was spending more time on sourcing for
key roles. Second most popular was more time communicating with key
candidates.
Automation and AI have great promise. I’ve seen claims of dozens of
hours saved from recruiting teams using different technologies powered by
AI. I just don’t want us to assume that people will gravitate towards
high-value activities once they have more time freed up in their schedules.
It’s not how we’re wired as humans, and if we’re not intentional, that time
will slip away and something else will take its place.

Employee sentiment analysis

Language is the fundamental currency for how people express them-
selves. The concept behind sentiment analysis is pretty simple: analyse
large amounts of text and categorize them by sentiment, or attitude.
The value is clear. By understanding the general mood of a popula-
tion of employees, whether globally, by department or location, a
company can better serve those workers and meet their needs. This is
done by blending natural language processing with machine learning.
64 ARTIFICIAL INTELLIGENCE FOR HR

Sentiment analysis relies on open, unstructured text from conversa-

tions, emails and other qualitative inputs. Additionally, machine
learning is used to train the system on words to look for that might
signify issues. If all workers in a specific function mention salary or
working hours in their communications or survey responses, then the
systems can flag those areas for review by an HR representative.
One example of a sentiment analysis platform is offered by a few
technology startups. The tools can analyse email data to determine
engagement, leadership, alignment and other elements. It’s common
for systems to look at engagement as a factor, but because these tools
plug into the email server data they can look more specifically at who
is emailing whom, whether they are having personal or work-focused
conversations, and more. While the systems are typically anonymous
and do not identify users directly, they can derive sentiment down to
the departmental level, giving employers the opportunity to address
issues and hot spots before they become a major problem.
A related approach is that taken by KeenCorp, which examines all
employee communication data on a continuous basis, identifying
sentiment across a range of use cases. For instance, it can determine
whether employees feel like they truly belong, supporting critical
diversity and equity initiatives. It can also be deployed during change
initiatives to examine how the workforce feels and where there may
be opportunities to improve engagement with critical stakeholders.
This is an example of inferred data as outlined in Chapter 2, where
the system infers or makes educated guesses based on existing data
without asking direct questions of the workforce. The benefit? It
doesn’t intrude on worker productivity or privacy at a personal level.

HOW AI CAN POSITIVELY INFLUENCE DIVERSITY AND INCLUSION

Recently, news broke that an AI-powered facial recognition technology

used by law enforcement was biased against, well, pretty much everyone
other than white men. This news hit the public like a slap in the face, but
it’s something that many of us have been watching closely in the hiring
 GENERAL AI APPLICATIONS WITHIN HCM 65

technology. Artificial intelligence as a technology isn’t good or bad – it just

‘is’. AI, at its core, is like a toddler. Over time it learns and can improve its
capabilities, but it isn’t smart enough to know the context and impact of its
decisions in most cases. If we use it properly and with care, we can improve
outcomes (including those related to diversity and inclusion). If we use it
carelessly, we can hamper our efforts and limit success both for our people
and for our organizations more broadly.
The negative effects of AI are very real. When you think about the
negative impacts of artificial intelligence, your mind inevitably goes to
something you’ve probably seen in a science fiction movie. Robots. Killer
AI. It’s a cliché, really, but the AI that I’m talking about today is less overt in
nature. Artificial intelligence algorithms are now being used to make
predictions for everything from child welfare support to recidivism rates. If
incorrect data is used, or if the algorithm has an underlying bias, then the
results could be disastrous for those on the receiving end of the decision.
Within the workplace, AI can cause flaws in recruiting decisions, causing
employers to avoid hiring qualified, yet underrepresented, candidates.
Amazon was courageous enough to come out publicly in 2018 and share its
own challenges with this process. Though some have disparaged the
company for its results, I believe sharing the cautionary tale is a laudable
act on the company’s part if it helps other firms realize the challenges that
may exist.
That said, there are some clear and definitive positive benefits of AI
when it comes to diversity and inclusion efforts. Let’s look at a few use
cases for how AI can help in the workplace:
●●
Uber uses an algorithm to set pay rates and schedule shifts for drivers,
which allows it to cut the gender pay gap by half of what it is in the
open market, improving pay equity for its more than two million drivers
across the globe.
●●
Unilever utilizes automated assessments and asynchronous video
interviews to find talented, diverse college graduates to join its team.
Moving away from a purely human-driven approach has increased
diversity and candidate satisfaction.
●●
Salesforce uses an external vendor technology that algorithmically pairs
employees with ‘guides’ to help coach them through difficult or
66 ARTIFICIAL INTELLIGENCE FOR HR

challenging situations. If we offered these resources to the entire

workforce, including diverse individuals, then we would see more
representation in the C-suite than we do today.

As you can see, the value in having an unbiased approach can lead to
better outcomes on a range of factors, including diversity. That’s because
machines are really great at certain things, but they’re terrible at others.
That’s where humans come in. The human skills of work, which are covered
extensively in the last chapter of this book, offer insights into how to
balance the scales between what algorithms can do and where humans
should play a role.

The American technology company UKG’s AI is similar in its approach.

The firm is doing its best to blend analytical intelligence with emotional
intelligence to get a fully accurate picture of what is going on in the
organization. The analysis then powers predictions, custom recom-
mendations and more. Instead of focusing on email data, the system
picks up other qualitative inputs from performance review comments,
social commentary on the company intranet and employee survey
feedback. The machine learning has been trained to the point where it
can understand general sentiment by extrapolating the user’s mood
from the text. For example, if someone types ‘The internet connection
is too slow at the office’, the system intuitively knows this is a negative
issue. Alternatively, if someone types ‘I have a great team’, the system
will associate this with positive feelings. While it is easy to look at
these examples and understand the intent from a human perspective,
imagine trying to do that tens of thousands of times across multiple
questions, workforce populations and demographics, and you start to
see just why it’s so important to use technology instead of humans to
manage the data analysis portion of this process.
A final example comes from employee feedback platform Culture
Amp. A certified B Corporation, the Melbourne-based software
provider’s goal is to help employers capture and act on employee
feedback. Head of Industry and Public Relations for Culture Amp,
Damon Klotz, told me, ‘Many companies have traditionally given
employees the illusion of listening. Instead of really listening to what
 GENERAL AI APPLICATIONS WITHIN HCM 67

needed to be done and then fixing it, they would collect survey data
and then sit on it, often letting the problems go unchecked over time.’
For the last few years since its inception, the company has focused
very heavily on the survey design and feedback collection processes.
It has a text analysis tool that offers employers insights into employee
sentiment. Unlike some text analysis tools that are trained using a
wide variety of unstructured data sets, Culture Amp feels like its tool
is better suited for employer purposes because the machine learning
algorithms were developed using thousands of employee feedback
comments gathered during the first few years of the company’s
­existence, not generic text models.

Skills as data in the modern enterprise

Increasingly, I’m seeing technology providers focusing on skills as a
sort of organizational currency to drive algorithms. We know that
data, and lots of it, is required to build machine learning models.
Skills offer a key inroad for providers as they have ties into every-
thing from hiring and compensation to performance and development.
There’s not a piece of the employee journey that doesn’t have some
sort of skills component woven in. When we combine that with the
fact that there are well-established repositories of skill data in use
around the globe (many based on the US O*Net framework), it
creates a valuable opportunity to create better and smarter practices
powered by skill data.
An example of this is Eightfold. The company leverages skill data
from millions of resumes to create a sort of skills hierarchy and rela-
tionship model. That powers many things within the platform, from
helping applicants see how competitive their skills are compared to
others that have applied for a job to giving recruiters a more concrete
sense of what skills a candidate actually has. In a practical sense,
Eightfold was selected from among 50 different companies by the US
Department of Labor for its 2020 Veterans’ Employment Challenge.7
The Veterans’ Employment and Training Service prepares US veter-
ans, service members and their spouses for meaningful careers,
provides them with employment resources and expertise, protects
68 ARTIFICIAL INTELLIGENCE FOR HR

their employment rights and promotes their employment opportuni-

ties. The system offers that audience opportunities to understand the
value of their skills and what career paths might align with them.
A different use case here is from Starmind. The company’s technol-
ogy examines ‘digital exhaust’ from workers across common
platforms like Microsoft SharePoint, Slack, Dropbox, Salesforce and
HR systems to determine what skills those people have. Customers
include enterprise firms like Bayer, Novartis and Accenture. When
someone in the company has a question relating to those skills,
Starmind can route the question directly to the right people, creating
a faster and more responsive organization. One interesting capability
is that those skills aren’t perpetual – if a user doesn’t demonstrate the
skill or keep it up to date, it falls off their profile, so the most current
experts are always front and centre.
Last but not least, Workday has also created some waves here with
its own Skills Cloud product. The system, which already houses
primary employee data, can leverage skills to help employees find
internal opportunities within the Workday talent marketplace.
Additionally, it’s worth noting that understanding what skills exist
within the organization is one of the most pressing issues for large
enterprises today. One client, Patagonia, nearly tripled its clarity into
the internal skills of its workforce using Workday Skills Insights.
This application of skill data across the organization has a tremen-
dous set of use cases, and it really exploded in 2020. The speed to
action that these tools enable will become a differentiator for organi-
zations as they look to start or continue with digital transformation
efforts.

Verbal communication and interaction

While email is one of the most commonly used tools for workplace
communication, there’s no substitute for stepping into someone’s
office for a quick conversation. Verbal communication always has
been and will continue to be a quick way to interact, and now the
second party in that interaction can be a machine, not just a person.
Virtually all of us have used a voice-based app on our devices to help
 GENERAL AI APPLICATIONS WITHIN HCM 69

us with searches, directions or other tasks – Apple’s Siri, Amazon’s

Alexa, Microsoft’s Cortana and Google’s Assistant all fill this need.
What’s more, voice interactivity has a variety of points that add to
the value the system can bring. For instance, most people can speak
faster than they can type. Additionally, those with mobility issues or
impairments might not be able to use a standard keyboard, but voice
interaction might offer them a way to interact with devices in a seam-
less manner. It also could increase the safety if someone can focus on
a task while speaking to a system without having to look at a screen
to type out a message. These are some of the reasons we have all
become enamoured with our various mobile assistants on our smart-
phones, and there’s no reason to believe this technology can’t also
add value to the HR environment.
‘How much vacation time do I have?’ In a demonstration at the
company’s annual user conference, Ceridian’s product team showed off
its voice capabilities and highlighted some of the coming improvements
in the system functionality. While examples like the vacation balance
request above require a user to initiate contact, other opportunities for
voice interactivity can take two paths: proactive responses and proac-
tive initiation. For instance, a proactive response to the question above
might be, ‘You have 65 hours of vacation. Would you like to schedule
some time off?’ On the other hand, a proactive initiation might be a
notification from the system based on an analysis of vacation balances
across the organization compared to the individual’s balance, a policy
capping vacation hours, or something else highly tailored. For example,
‘Alex, I noticed you have accrued 118 hours of vacation. Our policy
limits accruals to 120 hours, which means you will not receive any
additional hours for the coming pay periods unless you take some time
off. Would you like to schedule this now?’ In each instance, it’s almost
as if the employee has a coach guiding them through the conversation,
yet it’s nothing more than a voice-interactive algorithm programmed to
interact based on pre-set criteria or parameters.
The verbal frontier has much promise. One recent conversation
with a multinational firm was especially impressive. The company’s
HR team spent every Monday running reports for which employees
were on leave, investing several hours each week into this one process.
70 ARTIFICIAL INTELLIGENCE FOR HR

By partnering with a third party, they were able to integrate IBM’s

Watson for understanding verbal commands into the firm’s Workday
system, allowing users to run reports verbally instead of manually. It
was almost as if the users had an assistant to help them with their
tasks, because they could say, ‘Run leave report for EMEA region and
send it to Katy and Marcos’ and continue with their work. The system
would process the request in the background and complete the task
as assigned, minimizing the need for hands-on effort.
From a forward-thinking perspective, there remains an interesting
opportunity for incorporating devices like smart speakers into the
workplace and having them listen in on conversations, providing
insights after the fact. What if your speaker could listen in on the calls
your sales team makes and then give them ideas on how to improve
their results? Alternatively, what if your speaker could listen to
managers and then give them tips and ideas on how to improve their
coaching and feedback style? There are questions, and there should
be, about what might be recorded and how that information would
be used, but it’s a great example of the future prospects of what these
types of technologies can offer in terms of performance support in a
wide variety of use cases. I have yet to see a technology in the market
that is this advanced, but I know that we’re getting closer to these
types of capabilities in proactive technologies to help us all manage
our workloads, and our performance, more effectively.

CASE STUDY
Wearables: Mechanizing human networking and relationships

You’ve been invited to a networking meeting in your company, but you really don’t
want to go. After all, every time you attend one of these things you end up standing
by yourself in the corner checking out the latest buzz on your social media feeds on
your smartphone. But this time it’s different. You’re wearing a badge that will light up
when you get near someone with similar interests. As you walk through the group
navigating your way to the refreshments, you see your initials pop up on someone’s
badge nearby and you stop to greet them. After a moment or two you realize you
both have previous experience working for the same company in a different industry.
You spend some time getting to know this person and then continue on your journey
to get a drink, promising yourself you’ll reach out when you’re back in the office
because the individual’s area of expertise is one your team needs for an upcoming
 GENERAL AI APPLICATIONS WITHIN HCM 71

project. As you arrive at the table, your badge lights up with someone else’s initials,
and you turn around to greet them. You quickly realize that while your backgrounds
are different, you have a connection in your mutual love for children’s charities. That
starts the conversation, but before long you’re discussing work projects and how you
have overcome similar challenges in your respective roles. It’s not even ten minutes
into the event and you’ve already made two new connections you didn’t have before.
This story is one of the first examples I’ve seen of HR technology hardware
making its way into the employee experience by way of a wearable. We’re used
to software doing much of the work, and there’s definitely a software
component in this vendor’s solution, but the hardware element in the form of a
wearable device is intriguing because it is primarily a visual cueing system
powered by a software matching algorithm. This system was developed by
Chicago-based Proxfinity, and it works by having users answer a set of questions
prior to an event. Those questions could have a variety of designs, from looking
for ways to connect people with similar interests (as in the example above) or
looking for ways to connect people from teams and departments across the
organization to break down silos and improve communications.
What impressed me in particular was the analytics dashboard that employers
can review after an event. For example, if the system was used for an employee
onboarding event to help make sure executives socialize with new hires, the
team can analyse the reports after the fact to see if the executive team hung out
together during the event or if they were truly dispersed among the crowd and
making connections with the new hires. While this system isn’t yet AI-powered,
it’s a short step to incorporate some level of machine learning to support these
interactions. One component in a future artificially intelligent algorithm might
be helping to connect diverse individuals with key leaders, putting people
together without the benefit of the survey (using data from their employee
profile, performance goals and so on), or even matching high-performing
individuals with each other to spur innovation and idea generation.
As of today there are not many wearable use cases within the HR technology
realm beyond things like activity trackers that help with wellness initiatives. The
advent of more smart technologies will open up the opportunity for AI to make its
way into the hardware side of the equation, which offers an entirely different layer
of information for algorithms to factor into their calculations. When you think about
it, today algorithms are half-blind to the activities that occur in the workplace.
Sentiment analysis systems can see our virtual and electronic traffic and activities,
but our physical interactions are outside the scope of the system for the most part.
Creating this layer of data will enable artificially intelligent systems to look for even
better ways to help us improve the results we see with our human capital.
72 ARTIFICIAL INTELLIGENCE FOR HR

Automated chatbots
Bots. We’ve profiled a few of them already (and have several more to
go before we’re finished). It seems we can’t turn around today with-
out being inundated by yet another chatbot that fills a communication
gap in some form or fashion. That’s because they work, they are inex-
pensive compared to human labour, and they can be programmed to
fit a company’s verbal preferences, giving it a personalized feel.
Interestingly, one common complaint in some industry circles is that
chatbots should not be included in AI discussions because they are
typically nothing more than a library of responses that are used to
reply to conversational comments. Those detractors say that the
systems lack the machine learning components that get smarter with
every human interaction. However, chatbots are still automating key
parts of the business and its talent processes, which means for our
purposes they are worth exploring.
One point of value for chatbots is the opportunity for employers
to customize the language and wording so that the interactive assis-
tant mirrors the culture of the firm. For instance, an employer with a
strong focus on customer service might have the bot responses
programmed to use more language focusing on customer service or in
screening questions dedicated to that topic. Alternatively, if the
company is interested in being seen as having a creative or ‘fun’
culture, it can plan for the bot to crack jokes, build rapport, and
share glimpses of the culture through the various interactions with
candidates and employees.
Another example of chatbots supporting HR practices? Gifts.
Recently I had the opportunity to try out Eva, a bot that sends out
gifts to users, clients and candidates. From an HR perspective, it
could easily be used to send candidates a ‘thank you’ for their time,
but it could also be used for employees that need special recognition
for a job well done. The bot interacts with users via a simple chat
interface, quizzing them on their likes, dislikes and preferences. For
instance, during the conversation I mentioned I don’t drink, so Eva
didn’t send me a bottle of wine. Instead, the application sent me a
sampling of dark chocolate because my tastes tend to lean towards
the sweet side.
 GENERAL AI APPLICATIONS WITHIN HCM 73

On a more serious note, one area in which I believe bots can help
employers in an incredible way is by increasing HCM software adop-
tion rates. While it seems like a ‘meta’ discussion, it might be one of
the most undervalued points when it comes to bots supporting HR’s
needs. Why do adoption rates matter? Well, for those of us that focus
our efforts on analysing HR technology, we find that one of the top
reasons people select technology is usability. Essentially, if a piece of
software is intuitive and easy to use, it will be more valuable to the
organization. Bots help to make that interaction more seamless: it’s
easier to utilize a chat conversation than trying to find a help manual
or knowledgebase to learn how to use a piece of technology. This has
given rise to a plethora of tools in the HR space in recent years, from
Socrates.ai to MeBeBot and a dozen others.
Think about it this way: when a vendor sells a product to a buyer,
the value promise is based on full adoption of the system. For instance,
adoption of a recognition tool might promise up to 10 per cent reduc-
tions in regrettable employee turnover or a substantial increase in
employee engagement scores. However, if half of the employees or
managers refuse to use the tool, there’s no way it can meet the needs
of the business or generate the anticipated return on investment
(ROI). It’s like trying to create a new process to save resources and
having half the employees use the old process, negating the value of
the process change.
In instances where a firm has integrated bots into the product, such
as RamcoHCM, this can lead to higher adoption rates. This is cred-
ited in large part to the technology’s chatbots that are present on
Skype, Slack, Facebook Messenger, email and even text messaging.
Employees can interact with the HR system via any of these channels
according to the employee’s preference. For instance, if workers
prefer email, they can use Ramco’s email tools to handle HR-related
tasks. If SMS is the preferred communication tool, employees can get
pay slips and do chat-based interactions even on basic phones. While
smartphones are fairly ubiquitous, there are still areas of the world
and portions of the population where they are not fully utilized. Even
workers with ‘dumb’ phones that only have basic texting capabilities
(no internet access) can use these SMS-based tools to get work done.
74 ARTIFICIAL INTELLIGENCE FOR HR

Additionally, automation can be used to help users learn how to

use new software. WalkMe is a provider of technology that helps
businesses ease onboarding with new technology solutions. The
system uses a combination of AI and analytics to predict user behav-
iours and offer step-by-step assistance to increase engagement and
uptake with new systems. I constantly hear about the ‘user experi-
ence’ from technology providers and how they are trying to create a
positive one, and this example of having an assistant or tool to guide
you through the process as a new user of a system is a great example
of how to create an experience that users won’t soon forget.
At this point we’re seeing more and more uses for bots in almost
any area of HR that is heavily administrative or routine. We will
delve into other uses for bots in future chapters, as they are being
heavily adopted and leveraged in the talent acquisition process for
candidate engagement and screening. They are also being leveraged
to support employee engagement and recognition, an area where HR
often coaches managers. That sort of coaching practice can slip
through the cracks of a busy workload for many HR leaders, which
is why bots make the perfect assistant for those types of instances. In
the following chapters we will look at bots fit for a variety of purposes,
from candidate engagement and leader coaching to anonymous
complaints and common HR requests.

Facial recognition tools

In the last chapter I talked about the challenges of creating facial recog-
nition systems. As humans we can quickly scan and sort faces to learn
who is who, but algorithms have a harder time mapping the contours
of the face, creating unique maps of each individual person and account-
ing for hats, scarves, glasses or other obscuring items. Additionally, we
always assume that facial recognition works because it has a clear fron-
tal photograph to work with. In reality, photos can capture either side
of a face for someone looking away from the camera, which sometimes
makes identification challenging even to the human eye.
 GENERAL AI APPLICATIONS WITHIN HCM 75

Despite the challenges, several photo storage and sharing tools

today have the ability to recognize faces from a variety of photos
automatically, and this sort of advancement is also relevant to the
workplace. In a discussion with one firm in the HCM technology
industry, the representative showed me the company’s innovative
employee check-in system. This company uses a simple wall-mounted
tablet with facial recognition software in the front office of the build-
ing to capture the time you check in and check out for payroll
purposes. When an employee walks in, it quickly scans the person’s
face and projects a short welcome message on the wall monitor. This
helps not only with check-in time, logging their attendance records,
but with security and safety as well.
Facial recognition gets more visibility as a topic in Chapter 5 on
talent acquisition, because video hiring and interviewing solutions
are making increasing use of algorithms to process video content and
understand whether the interviewee is looking at the camera or read-
ing a script off-screen. By looking at the position of the eye and the
direction it is looking, the system can flag a person if it seems like
they are potentially cheating. We’ll also call it out appropriately in
the section on bias, as facial recognition systems can create issues
with hiring based on existing research.

THE FUTURE? FORECASTING HR ADVANCEMENTS BASED ON

CURRENT TECHNOLOGY

Retail workers will be able to point their mobile device’s camera at a specific
area of the clothing display and the system will recognize the product, offering
up a variety of helpful information to support the associate in real-time. The
device will bring up sales history for that item, including graphical
representations of pricing data over time. It will also display forecasted sales
based on current purchasing trends. In addition, the system can use that
predicted sales forecast to suggest sales discounts or pricing strategies to
maximize revenue.
76 ARTIFICIAL INTELLIGENCE FOR HR

While this sounds fantastical, it isn’t science fiction. It’s one of many
capabilities of Coleman, the AI tool that powers Infor’s systems across a
variety of industry verticals and its HCM platform. Infor is a multinational
enterprise software company headquartered in New York City. The AI
system is named for Katherine Coleman Johnson, a physicist and
mathematician whose brainpower and willpower helped man reach the
moon in the 1960s, and it is embedded throughout Infor’s suite of products
from healthcare and retail to human capital and supply chain management.
While this technology is amazing, the use case is focused on retail;
however, it’s easy to translate this visual recognition capability into the
HCM spectrum. For instance, what if a computer can recognize an
employee’s face in order to log them into an attendance tracking system?
Or maybe it analyses a person and determines the best training modality
for a particular employee’s learning style based on previous training and
performance? It might even suggest potential career paths by factoring in
the skills, training and aspirations of the worker.
While not all of these capabilities currently exist together, almost all of
them are out-of-the-box features for a variety of best of breed technology
vendors in the HCM technology industry. It’s exciting to think about how
these advances, taken as a whole, might shift how companies perceive,
manage and engage with the workforce.

Understanding and quantifying the value of AI systems

Despite the inherent excitement that comes from the capabilities these
systems offer, it’s important as a business leader to understand the value
of AI-based technologies. Is it more flash than substance? Do they really
do all the things they claim, and if so, what value does that add to the
business? Throughout this book discussions will turn to various value
points, but let’s look at some areas where potential value lies.

1 Productivity: According to Zoom.ai’s research, the most active

users leveraging the company’s automated assistant typically save
up to 25 hours a month, averaging more than $16,000 saved for an
employee with a $100,000 salary. Now, scale that across hundreds
 GENERAL AI APPLICATIONS WITHIN HCM 77

or thousands of corporate users, and it’s easy to see how this

becomes an incredible value proposition for the power of AI.
2 Adoption rates: It’s harder to pin a value on this, yet that doesn’t
mean it’s any less important. As covered in the section focusing on
chatbots, the idea that employees, managers and executives would
be more connected with and enabled by their firm’s technology is
a powerful one.
3 Performance of HR: Can we get more things accomplished if the
‘grunt work’ is being handled by algorithms and software? In
Table 3.1 you saw the diverse classifications for tasks: how many
non-urgent or non-important tasks might be captured and managed
by AI technologies, freeing up HR leaders to focus on the more
urgent and/or important activities?
4 Retention: While listing any specific number would be subjective,
the truth is that AI tools help with hiring practices, which could lead
to better employee fit and longer tenures. Additionally, the tools
supporting development and employee growth would naturally
increase satisfaction, adding yet another layer of possibility to the
retention figure. It’s too early to attribute a hard figure to this, but it
is logical to assume there is a connection.

One final note on value: HR is a profession that, in general, seems to

have challenges proving the value it can bring, and the investments in
HR technology (whether inclusive of AI or not) also face those same
challenges. That’s not necessarily a bad thing – we all need to be able
to prove what we bring to the table as business leaders. However,
what’s interesting is that executives clearly see the value HR technol-
ogy can bring. In one 2017 research study, we found that more
business executives outside HR thought their HR systems were stra-
tegic tools than the executives working within HR.8 This is
counterintuitive, but it’s also heartening to know that there is already
a measure of support for HR technology investment throughout the
C-suite. The conversation about the business value of HR technology
doesn’t stop there, but it’s certainly a good place to start.
78 ARTIFICIAL INTELLIGENCE FOR HR

KEY POINTS
●●
Investment into companies that develop AI-enabled HR software has
reached an all-time high and will most likely continue to climb.
●●
Jobs may or may not be impacted by artificial intelligence, but what we
do know is that there are incredibly valuable use cases spanning
everything from hiring and training to engagement and beyond.
●●
Verbal interaction, chatbots and facial recognition software offer just a
few of the foundational components of artificial intelligence technology
that feed up into the complex uses that span the HCM spectrum.

Notes
1 B Eubanks. Is AI better than humans at employee surveys? Lighthouse
Research & Advisory, 11 November 2019. https://lhra.io/blog/new-research-ai-
better-humans-employee-surveys/ (archived at https://perma.cc/G9AC-CVL5)
2 B Eubanks. Nearly half of 2017 HR tech funding events went to AI-focused
companies, Lighthouse Research & Advisory, 4 April 2018. http://lhra.io/blog/
nearly-half-2017-hr-tech-funding-events-ai-focused-companies/ (archived at
https://perma.cc/485C-EUDR)
3 J Abbruzzese. Americans are in denial about robots coming to take their jobs,
Mashable, 10 March 2016. https://mashable.com/2016/03/10/robots-are-
coming-to-take-our-jobs-just-not-my-job/#jvmDzzeHWuqF (archived at https://
perma.cc/WP3H-PG8P)
4 https://www2.deloitte.com/ca/en/pages/press-releases/articles/intelligence-
revolution.html (archived at https://perma.cc/6DXU-C78H)
5 R Kirkland. The CEO of Guardian Life on talent in an age of digital disruption,
McKinsey & Company, 20 March 2018. www.mckinsey.com/business-
functions/organization/our-insights/the-ceo-of-guardian-life-on-talent-in-
an-age-of-digital-disruption (archived at https://perma.cc/U2WE-KMV8)
6 C Gorman. Guess who wastes time at work, TLNT, 13 April 2017, www.tlnt.
com/guess-who-wastes-time-at-work/ (archived at https://perma.cc/BN25-
ME3S)
 GENERAL AI APPLICATIONS WITHIN HCM 79

7 B Gamble and G Vaught. US Department of Labor announces Eightfold.Ai

selected as grand prize recipient of Veteran Employment Challenge, US
Department of Labor, 18 September 2020. www.dol.gov/newsroom/releases/
vets/vets20200918 (archived at https://perma.cc/ER5Z-XH69)
8 B Eubanks. The business value of HR technology, Lighthouse Research &
Advisory, 1 August 2017. http://lhra.io/blog/business-value-hr-technology-
research-preview/ (archived at https://perma.cc/AT67-EP34
80

THIS PAGE IS INTENTIONALLY LEFT BLANK

 81

04

Core HR and workforce

management

In March 2019 a man passed away in his hospital bed in Fremont,

California. His family was near him at the end, but they had mixed
emotions. Just a few hours earlier, the hospital had sent a doctor to
provide a prognosis on the elderly gentleman for the family. When
the door opened, the family expected to be greeted by a specialist
with expertise in the particular affliction the patient was fighting.
Instead, a robot wheeled into the room, featuring a doctor on a video
screen. The doctor on the screen explained the terminal nature of the
prognosis, clarifying that the 78-year-old patient would soon die.1
After his death, the family’s reaction created a media firestorm. While
the family was satisfied with the medical care their loved one had
received, they were disturbed by the lack of humanity and compas-
sion in the delivery of this emotionally crushing news. They believed
that they were owed a personal, human response, not a computer-
delivered one.
This story is likely to play out in similar ways in the coming
years as we see more advancements in medicine, but it may bleed
over into the workplace as well. One of my repetitive phrases since
the first edition of this book came out several years ago is this: AI
will change how HR is delivered, but it won’t change the need for
the ‘human’ in human resources. Within this chapter we will look
at some of the more administrative aspects of HR, from payroll to
compliance to workforce management and more. Some would call
82 ARTIFICIAL INTELLIGENCE FOR HR

this mundane, but I’d caution that these aspects are anything but.
Pieces of the HR puzzle like pay and compliance form a critical
foundation for the other aspects of talent management and
employee experience that enable alignment with business strategy.

Core HR is core to the business

The client looked at me hopefully after explaining her company’s
current predicament. While still a fairly small company of just under
300 staff members, the complexities of managing time, attendance
and payroll were more than the HR team (a deeply passionate staff
of one) could handle. We were in the early planning stages of a tech-
nology selection and it was very apparent what this firm needed:
automation. Trying to juggle multiple systems, one of which was
hosted locally and faced regular downtime issues, was more than the
HR woman could bear. Every time the time tracking system went
down she had to get out the paper sheets and manually sort through
timecards to understand who worked on what project and when. In
addition, every data entry task she completed was duplicated
amongst several systems. And when she finally finished all the recon-
ciliations for time, attendance and leave, it was time to export the
data, upload into the payroll system and start the process all over
again for the next pay period.
On the other end of the spectrum, just a few days before I had been
talking with a 3,000-employee firm trying to move away from
in-house systems built by the company’s IT staff. What worked as a
home-grown system for 500 employees just didn’t scale up to thou-
sands of employees, and that led to a growing set of challenges and
frustrations. For instance, there was no official system of record.
Payroll had a system. Time and attendance tracking was totally sepa-
rate. Performance, learning and recruiting all had their own systems.
At any one time, getting something as simple as a headcount figure
was virtually impossible. Querying the respective systems would
return a variety of answers, each of them ‘correct’ in their own way.
 CORE HR AND WORKFORCE MANAGEMENT 83

Our goal was to consolidate and craft a plan for aligning the new,
integrated system with the firm’s growth plans and culture.
Both of these stories are not only true – they represent a wide
swathe of employers around the world. HR technology can be prob-
lematic because companies either buy targeted point solutions as they
grow, outpacing the capabilities they need to keep the firm opera-
tional, or they buy a large, complex system that sacrifices some
capabilities and features for integration and data integrity. In the
coming chapters we will look at the technology solutions that meet a
specific set of needs at virtually every company, but for now we’ll
look at the core HR component because it’s the most relevant to the
largest audience. After all, who doesn’t have to track who works and
pay those employees for their time?
Workforce management, payroll and risk management are some of
the fundamental, core functions of human resources. While they may
not be the most glamorous items on the ‘to do’ list for HR leaders,
these and other practices form the fundamental underpinnings of a
great function. Not sure if that’s true? Try letting payroll slip by a few
days and see what happens. Or the next time something pops up that
has compliance impacts, why not just let it slide for a few weeks and
watch the ramifications. Seriously, though, the need for companies to
deliver on these fundamental HR service offerings is critical to busi-
ness continuity. With the advent of artificial intelligence applications
in some of the fundamental aspects of how HR operates, companies
now have additional capabilities to support their needs.
While it’s somewhat vague, this concept of ‘core HR’ is going to
focus on a handful of specific aspects of the human resources func-
tion that don’t easily sit in other buckets, such as talent management,
learning and development, or talent acquisition. For instance, we will
examine how AI affects workforce management applications such as
scheduling shifts for workers. In addition, we will explore benefits
administration and the technologies that are simplifying that process
for employers. The conversation will also include compliance and
discrimination, a hot topic in the workplace made even hotter by
public firings for harassing behaviours in recent history.
84 ARTIFICIAL INTELLIGENCE FOR HR

Workforce management
Who’s scheduled to work tomorrow? What if I told you that there are
systems with the capability not only to help with the transactional
aspects of scheduling workers for upcoming shifts, but also to predict
the optimal scheduling mix of employees to deliver the most value to
the business and its customers? This is a reality in the healthcare field
with some scheduling products. The tools are designed to understand
the competencies of specific nurses and align them with the patient
needs as the schedule is being built, giving the patients access to the
right caregivers at the right time. Similarly, by matching up patients
with the unique competencies of a particular nurse, that practitioner
will feel validated. Everyone wants to know that their skills are
valued and important, and this type of workforce management tool
helps to do that for these workers.
The practical impact of expanding this ‘smart scheduling’ capabil-
ity into other areas is pretty astounding. For example, what if you
could predict in your retail stores which team members work best
together to create the best customer experience, increasing sales?
Instead of looking at scheduling as a transactional action, employers
would think more strategically about how to combine different skills,
personalities and competencies to create the right team for the right
challenge. Maybe one group is better suited for those intensely busy
times with high volume and high stress while another group is better
suited to the slower times because they can engage with customers
more deeply or develop creative merchandise displays that customers
love. This area will continue to mature as companies have the oppor-
tunity to connect more information sources like sales volume with
their workforce management and scheduling data.
While we’re exploring fundamentals of workforce management,
swapping shifts is one of those activities that can be challenging,
painful, annoying, or a combination of all three. Any worker that’s
traded shifts in the past knows that it can be difficult to find the right
person, contact them and make the change in the requisite system. At
its 2017 user conference, Ceridian demonstrated its Amazon Alexa
voice interactivity capabilities, one of which helps to solve this prob-
 CORE HR AND WORKFORCE MANAGEMENT 85

lem handily.2 In the demonstration of Dayforce Voice, an employee

goes through a short interactive dialogue to swap a shift with a
particular colleague and even schedule time away from work. While
those processes are not new or noteworthy, the idea of employees
being able to accomplish those tasks with a voice interface means it
will be even easier for workers to handle one of these common tasks.
Another firm working to manage the administrative needs of
scheduling and workforce management is UKG, formed from a 2020
merger of HR technology software giants Ultimate Software and
Kronos. One use case is around employees swapping shifts. In the
past, basic logic and rules governed the process for who you could
swap a shift with. However, the system is attempting to speed up the
process by highlighting the people you typically swap with over a
period of time and making smart recommendations based on those
actions. The algorithm essentially learns your behaviours, knows
who is most likely to accept that particular shift on that particular
day, and then it targets them with the request. Managers also get an
upgraded experience with visual tools that automatically highlight
which shift requests are okay to grant and which need human inter-
vention, potentially saving the manager time in navigating multiple
employee schedules. The platform boasts an automated chatbot that
interacts with employees to support interactive time off requests and
can automatically notify managers if dates are open. The bot doesn’t
replace the process – it helps to speed it up by having the schedules
and shifts all in one place and giving workers and managers an assis-
tant to facilitate the conversation. Because the chatbot is natively
integrated with Microsoft Teams, one of the most common enterprise
collaboration tools, and doesn’t require a separate interface, users
can more easily access the tool. In most cases, user adoption of new
technologies increases when vendors can find a way to get the inter-
active component into the flow of work, such as embedding it within
an already-used chat tool or putting links into the company’s intranet
for easy access.
While time tracking and attendance can seem like a mundane
concept in the bigger picture of HCM, the team at Replicon sees it as
anything but. In a conversation with some of the key executives at
86 ARTIFICIAL INTELLIGENCE FOR HR

the workforce management technology firm, they explained the firm’s

vision of ‘time intelligence’, or giving employers granular insights
into how the time is spent by their employees.3 For example, in a
200-person firm, business leaders have about 400,000 total work
hours a year to track and manage. While the business probably has
highly detailed accounting records of what money was spent on its
expenses, it’s probably less clear about how that aggregate set of
hours connects to actual objectives and outcomes. That happens
partly by being very clear about accuracy on the time tracking side of
the equation. One example of the firm’s innovative approach is in its
facial recognition time clocks. Workers simply walk up to the inter-
face briefly and the platform clocks them in for their shift. There are
no buttons or physical interactions required, and as a bonus there’s
no way for other workers to ‘buddy punch’ – the practice of clocking
in a friend before they arrive so they get full credit for the shift.

Payroll
After workers have been scheduled and employees have completed
their shifts, it’s time to make sure everyone gets paid. Employers
today have a variety of choices for running payroll, and in spite of the
powerful and relatively inexpensive options in the market, many
firms still use home-grown solutions. However, according to some
new research, this can be more costly than previously imagined.
Home-grown payroll systems have an error rate averaging just over
11 per cent, but employers using a third-party technology or service
see error rates around 6 per cent.4 The cost of payroll errors isn’t just
the time to correct the issue. You also need to factor in the very real
discussions the affected employee will have with their peers in the
workplace, potentially harming engagement and performance.
On top of that, new advancements in payroll technology allow for
on-demand pay. On-demand pay essentially allows a worker to access
their earned wages prior to payday. While this may not seem like a
major improvement, data from DailyPay shows that companies offer-
ing on-demand pay fill roles faster than those that don’t. In addition,
 CORE HR AND WORKFORCE MANAGEMENT 87

employees in some industries stay with their companies longer if they

have access to on-demand pay. In the world of personal finance,
people living from paycheck to paycheck is a common occurrence.
Tools like this allow employers to help the workforce balance their
personal needs. If someone has a medical emergency or a flat tyre
before payday, they can access those funds to avoid bank fees or
expensive short-term loans.
Anyone that has run payroll knows that the actual process itself is
ripe for disruption by AI: a highly repetitive, specialized process must
be run on a regular schedule with data that is significantly the same
over multiple periods of time. Oh, and let’s not forget the incredibly
high cost of errors within the payroll process. Forget about getting
the most productivity out of an employee with an incorrect pay slip –
you’ll be lucky to keep them engaged at all. At the same time, any HR
leader with a measure of credibility would hesitate to turn over the
payroll process to an algorithm. This unveils a critical decision point
that is going to increasingly become a common discussion in firms as
AI continues its advance: where do we turn over control to a computer,
and where do we retain it?
In this specific instance, it’s probably best to have AI process the
initial steps and have a human review the final version prior to
submission. This has several benefits. First, the most repetitive, time-
consuming aspects of running payroll come in the initial entries and
checks. Are all hours accounted for? Have employees used the correct
codes? Is overtime being calculated for the correct workers? What
about shift differentials and other pay modifications? And so on. The
initial steps of processing payroll are fairly routine and can include a
variety of these types of questions. Then, after that has been completed
and a draft payroll has been run, the payroll administrator is best
suited to the task of reviewing the data for completeness and accu-
racy. By the way, this process is very similar to what Immedis, a global
payroll provider, has developed in recent years. The company’s algo-
rithms can reference global compliance requirements for pay and
ensure that every payroll is compliant and accurate, something that
was left on the shoulder of payroll professionals in the past.
88 ARTIFICIAL INTELLIGENCE FOR HR

While this isn’t foolproof, it does bring to bear the best that humans
and software algorithms have to offer. A human can easily see if a
common deduction is incorrectly calculated or if a person on leave is
being paid accurately for their time off. At the same time, the algo-
rithm powering the payroll engine can do all kinds of instant
calculations that a human wouldn’t be suited for, such as determining
payroll tax rates for individual regions, cities and localities. For
instance, at one former employer our team was flummoxed by a pair
of employees living across the street from each other that seemingly
owed different types of payroll taxes. An algorithm using employee
residential data combined with a tax rules engine would quickly be
able to sort individuals into their proper municipalities and calculate
respective tax rates on the fly.
One reason pay and compensation is getting significant attention
in the media today is gender pay parity. Pay parity is all about ensur-
ing that women and men earn the same pay for the same work. It’s
easy to believe that in a world as advanced as ours, this kind of
concept should no longer be an issue. Yet the gender pay gap is still
alive and well. Sources vary but an estimate from the World Economic
Forum of the global pay gap puts the number at 32 per cent. Factoring
in countries around the world, we see that for every dollar a man
earns, a woman earns 68 cents.5 This number varies depending on the
geographic area, and only four regions are identified in the report as
having gender pay gaps less than 30 per cent: Western Europe (25 per
cent), North America (28 per cent), Central Europe and Eastern Asia
(29 per cent), and Latin America and the Caribbean (29.8 per cent).
The report looks at progress against pay parity year-over-year and
estimates it would take about 100 years to close the gaps for good at
the current rate of change.
One factor in this pay gap is psychological. Research shows that
women are generally less likely than men to ask for pay increases,
tout their performance and achievements and advocate for career
advancement. This natural phenomenon leads to women receiving
lower starting salaries and less in terms of ongoing pay increases. AI
could actually help with this by nudging managers or encouraging
the individuals themselves to recall their most valuable contributions
and highlight them during critical talent management discussions.
 CORE HR AND WORKFORCE MANAGEMENT 89

In a story highlighting the practical impacts, Salesforce has made

several very public pay adjustments in the last few years to correct
gender pay inequity.6 As a fast-growing company, the approach the
firm has taken is laudable. When the leaders realize there’s a pay gap,
they make moves to fix it. However, after the first adjustment back in
2015, the firm realized it didn’t fix the underlying factors with job
offers and promotions, which meant the issues reappeared after some
time. I highlight this because it’s a major challenge for all employers
to contend with, but there is hope that systems driven by AI would
enable us to minimize or eliminate this pay gap. In this case, what if
AI was continuously scanning the compensation rates across the
employee population and highlighting red flags or potential issues? In
the case of Salesforce, it could even scan for these kinds of imbal-
ances between men and women that would require massive pay-outs
to rectify. In the case study below I highlight an employer that uses an
algorithm to assign work tasks and pay rates along with the results
of the (ongoing) practice. It’s an interesting story, not just because it
shows the usage of an algorithm in action but because the results of
the activity were still not exactly what the employer expected.

CASE STUDY
Does AI ensure pay parity?

As we’ve discussed, one benefit of artificial intelligence is a system that makes

decisions without bias or regard for someone’s gender. Therefore, if you could
design a system that schedules work shifts and pay rates based on a blind
algorithm that does not factor gender into the decision, you would logically
expect to find that men and women earn the same in such a system, correct?
But what if I told you this isn’t the case? There’s an employer that exists in
markets around the globe, and in a recent analysis by economists from Stanford
and the University of Chicago they found that in spite of this highly automated,
gender-blind algorithm that sets pay rates and assigns work in real-time, men
still out-earn women. This employer, if you’re curious, is Uber. In an analysis
released in 2018, several economists looked at the transactions that occurred in
the system to understand whether there was a pay gap.7 Transparently, one of
the economists admitted that he fully expected to see little to no gap in pay
90 ARTIFICIAL INTELLIGENCE FOR HR

because of the structure of the system. Again, we all logically expect this. Yet the
conclusions of the analysis are equally logical, if a little confounding for those of
us that had hoped to find a mechanism for eliminating the gender pay gap.
The gender pay gap for Uber drivers is 7 per cent, and the gap is explained by
three key factors. First, experience accounts for about 33 per cent of the gap. To
put it simply, drivers with more trips earn more than drivers with fewer trips.
Because men have longer tenures on the Uber platform, on average, they reap
the benefits of this. Drivers who have taken more than 2,500 trips earn an
average of $3 per hour more than those drivers with less than 100 trips. Second,
driving speed accounts for about 50 per cent of the difference. Men drive faster,
which means they can get more pickups per hour. Men drive marginally faster
on average in the broader population as well. Third, variations in work times and
routes make up the remaining 17 per cent. Men take on shifts during higher
surge times and in surge-friendly locations, leading to higher hourly earnings.
Surges occur when there is higher demand for drivers than the available supply,
which pushes up the cost per drive for passengers.
It’s worthwhile to note the gap is better than what we’d find in the open
market, and it’s explainable, too. What is important is that pay assignments are
equal. Men and women that drive the same route at the same time earn the same
pay. In that respect the algorithm really is levelling the playing field. However, in
terms of hourly wages, men are earning slightly more because they have more
experience, faster driving speeds and more lucrative routes/pickups. The logical
conclusion after reviewing this information is that unequal results don’t
necessarily point to unequal treatment at the outset. In this case the algorithm
worked exactly as advertised, and human unpredictability explains the rest.

Benefits administration
‘Alexa, what amount of my deductible has been met?’
We have already discussed the importance of voice-interactive
systems and the value they bring to users, but they also pose a new
frontier for benefits conversations. In an interview, benefits executive
Richard Silberstein pointed out that the benefits profession is on the
verge of being disrupted by technologies enabled by artificial intelli-
gence.8 His belief is that the technology will allow us to go beyond
the traditional ‘open enrolment meeting’ focal point where a human
is required to explain and explore benefit options to individuals in
 CORE HR AND WORKFORCE MANAGEMENT 91

the workplace. While that human touch is still a part of the process
and will not completely go away, the value of having more personal-
ized discussions with individual workers cannot be overlooked.
Even if we don’t always like to admit it, each of us wants to feel
special. We want to be treated uniquely. One area of the business and
how employee interact that demands a level of personalization is in
benefits selection and administration. Representing up to 40 per cent
of an employee’s cost on top of their salary, benefits make up a major
line item in every company’s budget.9 That’s precisely why we need to
be sure that our approach to benefits ensures that every single
employee gets the most beneficial option that fits their unique needs,
and one of the best ways to do that at a micro or macro level is
through communication. The use cases for employee benefits commu-
nications are fairly broad, and I’ve seen personally what happens
when you throw all of the options and information at all of the
people: they have some level of paralysis due to the information over-
load and either select a default option or just maintain course, even if
other options might be better suited to their needs and lifestyle. But
we can change this, and it isn’t difficult. For instance, maybe you
define your communications based around employee age group:
●●
Baby boomers – Offering childcare options to baby boomers is a
missed opportunity for impact. Instead, offering prescription drug
benefits or long-term care coverage would be more appealing for
that particular audience.
●●
Gen Xers – Dependant care coverage, retirement planning
assistance, elder care assistance, or college planning for young
children.
●●
Millennials – Student loan repayment programmes, financial
wellness training/assistance or coaching for first-time healthcare
insurance buyers.

When HR leaders stop and think logically about this, it makes perfect
sense. Different parts of the employee population have different needs
at different times. This idea of communicating proactively is some-
thing that US healthcare firm HealthJoy weaves into its approach to
92 ARTIFICIAL INTELLIGENCE FOR HR

the market. The company’s bot will reach out to members at strategic
times of the year to remind them to get a flu shot. However, unlike a
generic message or text reminder, the system also automatically scans
insurance information to find the best local provider. For those with
company-sponsored insurance, it advises users on which providers
are covered by the insurance plan, and for those with national health-
care offerings it might simply suggest a local provider based on
customer reviews and ratings. This kind of personalized, tailored
outreach would have never happened if it required a human to
manage the process, but allowing an algorithm to run this costs the
company virtually nothing and may reduce costs and expenses asso-
ciated with presenteeism (coming to work while sick).
One AI scientist even envisions a future where some of the common
employee assistance programmes or even well-being benefits are
being handled by an algorithm.10 Employee assistance programme
plans are notorious for being rolled out by companies with much
fanfare, only to have little to no adoption among the employee popu-
lation. This scientist believes chatbots could serve as basic therapists,
trainers or assistants to help bridge the gap between employee needs
and a human response. On the wellness side of the equation, instead
of paying for personal trainers for workers, the bot could easily serve
as a coach for basic nutrition advice, exercise and more. The cost for
the employer would be virtually nothing per interaction, but research
shows that increases in employee mental and physical health (and the
accompanying reductions in undue stress) can improve their work-
place performance.11
For companies, there are other opportunities on the benefits front
that might be challenging to manage with human labour. Employers
with self-funded insurance plans might use algorithms to target recur-
ring medical issues or trends that are driving up plan costs, seeking
alternatives to reduce those specific types of issues as needed. Or it
could be about making sure people actually adopt the programmes
and benefits the company is already paying for, keeping their satisfac-
tion high with their offerings and making sure they maintain their
health to the degree possible. It’s no stretch of the imagination to
 CORE HR AND WORKFORCE MANAGEMENT 93

envisage bots helping with that adoption rate similar to how we

discussed bots helping with technology adoption rates earlier in this
book. Instead of giving a worker 200 pages of documents and expect-
ing the person to make a choice about what benefits fit their needs, a
bot could assist each person with their questions, unique circum-
stances and budget to help them find the best solution. On top of
that, it could follow up throughout the year to support adoption and
ongoing usage of the selected benefits. For example:

‘Mary, if you remember, you signed up for the flexible spending

account to save money for medical expenses but you haven’t yet taken
advantage of that. Those funds expire at the end of the year so be sure
to use your flex spending debit card when you’re paying for covered
expenses like co-pays and deductibles. If you don’t have your card, I can
replace that for you or I can get you the reimbursement form if that’s
easier for you to use.’

Again: this would cost nothing but would potentially increase Mary’s
satisfaction with her benefits because she is able to take full advan-
tage of what she and the company are paying for. Alternatively, the
system could work the same way but in a reverse or look-back
scenario.

‘Jamie, I see that you spent $1,700 out of pocket this year for medical
expenses. Did you know that could have been tax free under one of our
medical savings account options? If you like, I can help you learn more
about those options for the coming plan year.’

David Contorno, a benefits broker, has seen the implementation of AI

into programs like telemedicine increase utilization twenty-five fold,
simply because the system can be more personalized and interactive
with each individual person.12 If you’re not familiar, telemedicine is
an option where employees can call or video chat with a doctor
instead of having to visit a doctor’s office for some routine issues: ear
infections, pink eye, sinus issues, and so on. The cost is dramatically
lower for both the employee and the employer than an actual doctor’s
visit, and the physician can even prescribe antibiotics and other medi-
cations without requiring an in-person visit. These programs are
94 ARTIFICIAL INTELLIGENCE FOR HR

incredibly efficient uses of benefits resources, but they don’t always

have high adoption rates by employees because they forget about it
or they are unsure how to use the benefit. The cost saved is approxi-
mately $100 per virtual session or use versus visiting a doctor in
person, even if we don’t account for the issues associated with driving
to the office and waiting in an overcrowded room, potentially expos-
ing ourselves to other germs in the process.13 Now, imagine expanding
those cost savings across your employee population thanks to the
value and scalability of AI technologies and you’ll start to see the
vision Contorno has seen for the companies he works with.

Employee self-service
The value promise of employee self-service has long been that work-
ers can handle some of their own tasks without having to involve
HR in the process, giving HR teams more time to dedicate to non-
administrative, or strategic, actions. If that sounds familiar, it
should – it is the same overall value promise of AI for HR leaders as
discussed earlier in this book.
When I think about the kinds of questions I answered as an HR
specialist in my early days, a significant portion of them were repeated
over and over again. How do I change a beneficiary? How do I roll
over a retirement plan? How do I check my vacation leave balance?
These kinds of questions were repeated on an almost daily basis by
employees, and a tool that would have enabled them to get those
answers more quickly and consistently while saving me time to do
other activities would have been incredibly valuable.
The concept of self-service has been around for as long as HR
teams have adopted technologies that allow user inputs from outside
the HR team, but don’t assume that this is a common staple of all
employers, even large ones. In one Infor case study exploring the use
of technology at Pilot Flying J, an operator of travel centres and rest
stops, the company’s CIO Mike Rogers discussed the firm’s adoption
of a cloud-based HR technology suite.14 Rogers explained that, until
recently, his company’s 27,000 employees had been using paper
 CORE HR AND WORKFORCE MANAGEMENT 95

forms and fax machines to handle even simple employee address

changes. This represented a major frustration for workers, and in a
high-turnover industry even the smallest details could lead to a more
broadly negative employee experience. The company’s goal was to
reduce turnover by half by adopting new HR systems to reduce that
level of friction.
While anecdotal, this represents a common challenge for employ-
ers worldwide. How do we serve the needs of a population without
completely overwhelming our HR infrastructure and resources? This
is a perfect opportunity for bots to support employee needs, extend-
ing the capabilities of the existing human resources function and
enabling a more consistent, high-touch employee experience. When it
comes to the employee experience, there is a variety of lenses to look
through: cultural, physical and even technological. One company
that focuses on creating a frictionless employee experience from an
IT perspective is Credit Suisse. The financial services firm headquar-
tered in Zurich has nearly 50,000 employees, plus a special one called
Amelia. When workers have issues with passwords, logging into their
systems and other routine issues, they reach out to Amelia for help.
Just like a normal IT help desk, Amelia can work with employees to
resolve their issue and get them back to work. However, Amelia has
the benefit of being able to serve multiple employees at the same time,
regardless of time zones, and she doesn’t require overtime pay. Why?
Because Amelia is a chatbot, and this bot is designed to scale up
support for common issues to keep employees from being frustrated
by minor IT annoyances, helping the internal IT team focus on more
important issues than password resets and other similar requests.
This type of employee self-service, where workers partner with an
automated assistant, is becoming increasingly common, whether in
IT, in HR or in other areas of the business. Smart HR teams will look
at these assistants not just as a tool designed to solve a narrow set of
problems, but as an additional resource on the team. Just like junior
staffers often have tasks delegated to them that are not worth the
time investment for more senior workers, bots and automated assis-
tants fill that same role for an HR team.
96 ARTIFICIAL INTELLIGENCE FOR HR

Spoke is one firm that offers these automated assistants for human
resources teams. The CEO, Jay Srinivasan, believes that these tools
have the ability to support not just enterprise HR teams but small
departments-of-one as well.15 Having run HR functions as a small
team of me, myself and I in the early days of my career, I know all too
well how easy it is to get bogged down in the details and never take on
the more strategic, forward-looking aspects of the job. Srinivasan says
that employee self-service tools are commonly marketed towards
larger firms, but small companies have the same basic needs around
task management, ticketing and support. Spoke’s team has seen that
nearly half of all questions that are submitted to HR teams could be
answered immediately and automatically by an algorithm, requiring
no downtime or lost productivity for the HR team. Another interesting
point about Spoke is that it has adopted a Slack-like pricing model,
where customers pay only for the active users in the system each month,
not a set number of seats or licences. Why does that matter? Because
this is a major incentive for providers like Spoke to make the product
incredibly easy to use for employees, driving adoption and usage.
Another example of an employer that uses HR self-service chatbots
is E W Scripps. Founded in 1878, the firm’s 4,000 employees rely on an
automated tool from Socrates.ai to support their daily needs. Employees
that use the chat interface can get access to multiple systems at one
time because the chatbot is connected with the company’s HR case
management, workforce management, benefits and human resource
information systems. If an employee needed answers from several
places, that could take some time to log in to each system, navigate the
interface and find the solution. However, since the bot is integrated
with multiple systems at once, the employees can get that information
at their fingertips using natural language queries instead of trying to
figure out where to navigate in a system they don’t use very often.

Diversity and harassment

In a world shaped by the #MeToo movement, demands for racial
parity and LGBTQ communities, there is an incredibly heightened
 CORE HR AND WORKFORCE MANAGEMENT 97

awareness of diversity, harassment and compliance. Interestingly,

both of the companies we’ll examine below were co-founded by
women and both have goals to reduce bias and discrimination in the
workplace.
The first is Spot, co-founded by Julia Shaw, a criminal psychologist
and memory expert at University College London. The tool uses a
chatbot to interview workers about their harassment experience and
then can submit the anonymous report to a trusted person within the
company. Interestingly, this can be used to help circumvent HR within
the process, despite HR often being the default ‘first stop’ in the
reporting mechanism for any discrimination or harassment issues.
This is intentionally designed that way for a few reasons: first, it
avoids the discomfort of being identified and making a complaint in
person, and second, the unemotional interactive chatbot may help to
draw out more accurate information from complainants as they
recount their experiences.
This concept is a valuable one because it takes advantage of the
flexibility and always-on nature of chatbots and combines it with the
need to approach a situation without any preconceived notions or
defensiveness. In the highly publicized 2017 events that led to uproars
about gender discrimination in the workplace, several of the stories
included a variation of ‘I went to HR but nothing was done about my
situation’. If this type of tool can help to alleviate harassment for
even a single person, it should be considered a success. However, the
adoption and usage could surprise those of us that know how often
harassment allegations typically go unmentioned. According to one
interview, Shaw said that within just a few weeks of making the tool
freely available to the public, it had been used more than 200 times.16
The second tool, targeting bias in the workplace, is doing so in a
slightly different manner. Joonko, named for Junko Tabei, the first
woman to summit Mount Everest in 1975, is a real-time diversity
coach that utilizes machine learning and natural language processing
to identify and uncover biases, whether conscious or unconscious. In
practical terms, the system plugs into existing HR software to get a
snapshot of the workforce as it currently stands. It then can start
making recommendations based on areas to improve. For example,
98 ARTIFICIAL INTELLIGENCE FOR HR

maybe recruiting pipelines are not very diverse, which will have a
cumulative negative effect on company diversity in the long run.
While we’ll explore other types of recruiting tools in the next chapter,
this is one example of how diversity blends over into other areas of
the function and the business. On the company’s website, visitors can
see a short demonstration of how Joonko’s AI can automatically flag
various words and phrases in a Slack chat conversation, encouraging
users to find other, more suitable terms to share their thoughts. It’s
early yet for this particular startup, but as other use cases emerge one
can only imagine that these types of tools will find a greater foothold
in the workplace environment.

AI-ENABLED WORKFORCE PLANNING

One of the more complex elements that spans workforce management,

talent analytics and business impact is workforce planning. Many
employers struggle with this process. Even if they have a handle on current
hiring needs, few firms have the practices and structure in place to be able
to predict demand for specific types of talent and skills into the future.
However, some firms are seeing AI give them the edge in this process.17 A
powerful tool that many enterprise organizations have leveraged in this
area was developed by Visier. The company’s analytics capabilities are
award-winning and have led to many companies finally getting clarity on
what’s happening within the business.
By analysing a variety of supply sources for talent, businesses can
predict where their talent is most likely to come from. Consider for a
moment where your candidates might be found: social networking sites,
applicant tracking systems, university databases, and more. Employers can
pool these various inputs to create a massive database of information
about employment trends and sources. They can then use machine learning
algorithms to develop workforce planning models. There may be trends in
sources that aren’t obvious to the naked eye, but an algorithm might be
able to shed light on those with ease. The other side of the equation is
around talent and skill demand. How does the employer know what it will
need? Part of this comes from analysing business context and goals. If the
company plans to pursue a new type of technology development, those
 CORE HR AND WORKFORCE MANAGEMENT 99

skills that apply to that technology field will become increasingly prioritized
in the workforce planning model.
Matching up the supply and demand sounds simple, but the reality is it
takes a significant amount of work. The positive side to this conversation is
that AI can take away much of the struggle and stress that this process has
historically involved. Instead of relying on a spreadsheet and pivot tables to
analyse reams of data, employers can leverage smart algorithms to identify
opportunities and trends that exist to hire the right people at the right time
for the right task.

What I hope you’re left with here is the idea that some of the more
mundane and/or unpleasant administrative functions of HR are at
the cusp of automation. And the promise of automation in these
areas is greater time availability for important tasks like hiring, train-
ing and engaging the workforce. Because of the transactional nature
of the tasks and activities covered in this chapter, combined with the
high cost of errors, these will be some of the first areas to be signifi-
cantly impacted by artificial intelligence technologies; however, they
won’t be the last.

KEY POINTS
●
Payroll, workforce management and benefits administration all share
common characteristics: highly routine, high volume and high cost of
errors. This makes them ripe for disruption by artificial intelligence.
●
Offloading tasks from HR teams to the employees themselves using
self-service and chatbots can free up HR staff to focus on more critical
aspects of the business.
●
Employers can’t focus on actions like hiring and training until they have
ironed out the issues around core HR and workforce management. Only
then can they move on to more strategic activities.
100 ARTIFICIAL INTELLIGENCE FOR HR

Notes
1 D Andone and A Moshtaghian. A doctor in California appeared via video link
to tell a patient he was going to die. The man’s family is upset, CNN Health,
10 March 2019. www.cnn.com/2019/03/10/health/patient-dies-robot-doctor/
index.html (archived at https://perma.cc/M2C9-HNHS)
2 B Sommer. Insights from Ceridian Insights 2017, Diginomica, 15 October
2017. https://diginomica.com/insights-ceridian-insights (archived at https://
perma.cc/AU3J-FWZW)
3 Replicon. Time intelligence, 2021. www.replicon.com/time-intelligence/
(archived at https://perma.cc/ZE2M-9KGM)
4 M Lombardi. Why I’m passionate about payroll, HR Tech Insiders, 23 July
2018. https://blog.hrtechnologyconference.com/why-im-passionate-about-
payroll (archived at https://perma.cc/7LZS-K3TN)
5 WEF. The Global Gender Gap Report 2017, 2017. www3.weforum.org/docs/
WEF_GGGR_2017.pdf (archived at https://perma.cc/LBC6-D8XC)
6 M Dickey. Salesforce spent another $2.7 million to adjust pay gaps related to
race and gender, Tech Crunch, 17 April 2018. https://techcrunch.com/2018/
04/17/salesforce-spent-another-2-7-million-to-adjust-pay-gaps-related-to-race-
and-gender/ (archived at https://perma.cc/V2EV-8JWC)
7 C Cook, R Diamond, J Hall, J List and P Oyer. The gender earnings gap in the
gig economy: Evidence from over a million rideshare drivers, Stanford, May
2020. https://web.stanford.edu/~diamondr/UberPayGap.pdf (archived at
https://perma.cc/7QCS-SDLE)
8 P Albinus. AI poised to transform open enrollments, Employment Benefit
News, 18 January 2018. www.benefitnews.com/news/ai-poised-to-transform-
open-enrollments (archived at https://perma.cc/9NN7-Q8PG)
9 J Hadzima. How much does an employee cost? MIT, 2005. http://web.mit.
edu/e-club/hadzima/how-much-does-an-employee-cost.html (archived at
https://perma.cc/G7VV-6TNA)
10 B Shutan. AI’s brave new world of insurance apps, Employee Benefits Adviser,
1 October 2017. www.employeebenefitadviser.com/news/ais-brave-new-world-
of-insurance-apps (archived at https://perma.cc/X7VJ-DSEV)
11 F Gino. Are you too stressed to be productive? Or not stressed enough?
Harvard Business Review, 14 April 2016. https://hbr.org/2016/04/are-you-too-
stressed-to-be-productive-or-not-stressed-enough (archived at https://perma.cc/
FR8R-2MDL)
12 R Ramos. AI is making employee benefits a little more beneficial, Venture Beat,
24 October 2017. https://venturebeat.com/2017/10/24/ai-is-making-employee-
benefits-a-little-more-beneficial/ (archived at https://perma.cc/9FXP-ATXR)
 CORE HR AND WORKFORCE MANAGEMENT 101

13 K Wike. Telehealth costs up to $100 less than office visits, Health IT Outcomes,
24 December 2014. www.healthitoutcomes.com/doc/telehealth-costs-up-to-less-
than-office-visits-0001 (archived at https://perma.cc/E55B-N9TZ)
14 M Rogers. 5 reasons pilot chose Infor CloudSuite HCM, Infor, 2021. https://
pages.infor.com/hfe-hcm-case-study-pilot-flying-j-cis.html?cid=NA-NA-HCA-
US-HCM-0716-FY17-HCM-Assets-WWCS-43525 (archived at https://perma.
cc/M7HY-HMFD)
15 I Banerjee. Demystifying ESS to help HR succeed: In conversation with Jay
Srinivasan, CEO at Spoke, HR Technologist, 13 February 2018. www.
hrtechnologist.com/interviews/employee-self-service/interview-with-jay-
srinivasan-ceo-at-spoke/ (archived at https://perma.cc/MS7X-XF2L)
16 P Olson. This chatbot is helping people track harassment at work, Forbes, 2
March 2018. www.forbes.com/sites/parmyolson/2018/03/02/chatbot-spot-
sexual-harassment-ai/#1bff2fc84a51 (archived at https://perma.cc/H6D4-
NFB2)
17 D D’Onforio. How a telecoms giant is using AI to predict its future workforce
needs, HR Technologist, 13 February 2018. www.hrtechnologist.com/articles/
recruitment-onboarding/how-a-telecoms-giant-is-using-ai-to-predict-its-future-
workforce-needs/ (archived at https://perma.cc/4HKT-SRJM)
102

THIS PAGE IS INTENTIONALLY LEFT BLANK

 103

05

Talent acquisition

The date is March 2017. The place is Austin, Texas. The audience is
on the edge of their seats waiting to hear the results. Who will win –
the top sourcing experts or the algorithm?
The SourceCon Grandmaster Challenge is well known in the sourc-
ing industry, but others outside the field might not be familiar with
this competition. The contest for 2017 was fairly straightforward:
download a folder containing three job descriptions, download a
batch of 5,000 resumes, and examine the resumes to determine who
was actually sourced, interviewed and hired. In other words, contest-
ants would review skills from the resumes and predict which of the
individuals made it through an actual hiring process. Points were
given in each case for selecting the right resumes and their classifica-
tions (interviewed, hired, etc.). At the same time as the humans, a piece
of AI-driven technology from Brilent was also attempting to solve the
riddle. Spoiler alert: the humans won, but it was a tempered victory.
We’ll talk more about the process, the results and the practical
implications in the sourcing section below, but this epic battle is the
perfect lead into the conversation about how AI tools fit into the
recruiting function.
Talent acquisition is often practised by HR professionals who have
a measure of innovative thinking and a healthy appreciation for
getting results. They also have low rules orientation – this essentially
means they are less interested in following a policy manual and
instead seek autonomy in how they accomplish their work. For
example, most companies don’t have a recruiting ‘manual’ that tells
104 ARTIFICIAL INTELLIGENCE FOR HR

how and where to find candidates. Instead, recruiters are often a little
bit like scientists tweaking, testing and experimenting with their tools
and skills to find the right candidate for the job.
The above-referenced competition is a great example of how
this plays out in real life. This combination of factors has led to a
proliferation of recruiting applications using artificial intelligence
technologies. Because the role of a recruiter is more routine than,
say, an HR business partner, the recruiting function is more easily
supported by automation technologies. Consider the typical work-
flow for a job: the requisition opens, candidates apply, candidates
are assessed, candidates are screened and somewhere down the line
the candidate actually talks to a recruiter or hiring manager. All of
those early steps are opportunities to leverage chatbots, automated
screening tools and other systems to scale the interactions with
candidates.
On top of that, the hiring process generates an immense amount of
data, one of the other core foundational underpinnings discussed earlier
in this book. Because data privacy and security are such critical points
for many employers, these topics are highlighted in detail in Chapter 8.
If a company hires 500 workers per year and receives 5,000 job appli-
cations, those 500 that were chosen as well as the 4,500 that were not
create some opportunities to understand trends and gather insights on
the types of workers and skills a company desires. This ratio of appli-
cants to openings is a conservative figure, as some estimates put the
average number of applications per job opening at 250.1 Regardless, the
data tells a compelling story about how companies hire and can shed
light on what can be done to improve the process.

TALENT ACQUISITION CHATBOT USE CASES

Chatbots are perhaps one of the most well-known types of tools that use
natural language processing and AI to support hiring. The most common
use cases help take a casual website visitor down the path to an interview
 TALENT ACQUISITION 105

and possibly even beyond, depending on the provider and functionality.

Below are some of the common capabilities:
●●
Capture/apply: Capture traffic and converse with candidates that visit the
career site, answering questions and ultimately converting them from
visitors to applicants.
●●
Screening: Ask basic qualifier questions of candidates to identify job
matches and help slot candidates into the appropriate positions.
●●
Scheduling: Serve as intermediary between recruiters and applicants to
schedule interviews. Includes email, SMS and other commonly used
messaging channels.
●●
Other: This is a ‘catch-all’ but provides a bucket in which to place
chatbots that have additional functionality. For instance, if a bot does
post-hire check-ins to determine satisfaction, this would be classified as
‘other’.

While the functionality can vary a bit, the commonality among each of
these tasks is that they are currently usually done by humans but are easily
automated. Any routine task that follows a narrowly defined path is an ideal
candidate for automation, whether in recruiting, accounting, marketing or
any other area of the business.

The pressure to interact with and support each applicant has also
been driven by a shift in demand in recent years. Today, candidates
expect greater responsiveness and transparency from employers than
ever before. In one of the world’s leading candidate experience
research publications, the 2017 Talent Board North American
Candidate Experience Research Report, the authors recommended
giving candidates similar experiences to consumer shopping where
they can see their progress or understand how long the process might
take before undertaking the application steps.2 While technologies
have given companies more opportunities to scale their recruiting
practices, they have also been known to incense candidates who
106 ARTIFICIAL INTELLIGENCE FOR HR

apply to many jobs without a response. In reality it takes only a few

minutes to let candidates know they are not chosen for a position,
but many employers still don’t utilize that functionality in their appli-
cant tracking systems. It’s rare to go to a recruiting event and not
hear about the ‘black hole’ – the descriptor given by many to the
applicant tracking systems where resumes are dumped but no
communication is ever returned. This greater desire for transparency
means the stakes are higher than ever for employers that want to
attract and hire the right talent.
Additionally, a greater focus on diversity in today’s business
climate increases the perceived value of automated systems. One
use case is resume blinding, where the name or other identifying
information is removed from the resume to prevent conscious or
unconscious bias. In one study, applicants with white-sounding
names like ‘Emily’ or ‘Greg’ received nearly 50 per cent more call-
backs than candidates with black-sounding names like ‘Lakisha’ or
‘Jamal’.3 In the context of the study, the researchers found that
having a white-sounding name is worth as much as eight years of
work experience. The practical implication of this for purposes of
the AI discussion is that a piece of technology should, theoretically,
reduce the bias in the process by focusing on work experience,
skills or other non-discriminatory factors. As we’ll see in some of
the examples within this chapter, some of the technology providers
in the space are also taking alternative approaches to reduce any
bias in the process from the earliest resume screens all the way
through to the interview process.
One other comment on diversity that has been particularly
intriguing in the research process for this book: many of the startup
firms in the HR technology industry are founded by women, and
while not all of them try to tackle the same problem or step in the
process, several of these businesses have a distinct approach for
minimizing bias and discrimination. The list is long and varied, but
I’ll give you a glimpse in Table 5.1 into a sampling of the types of
tools that are being developed (several of which will be explored in
this chapter).
 TALENT ACQUISITION 107

TABLE 5.1 A sampling of female HR technology founders with AI-based recruiting

technology

Name Company Focus

Athena Karp HiredScore Screening and matching candidates in an unbiased way

Kristen Hamilton Koru Candidate assessments

Frida Polli Pymetrics Gamified candidate assessments

Kieran Snyder Textio Augmented writing tool to minimize bias in job ads

Laura Mather Talent Sonar Structured interviews to remove hiring bias

Stacey Chapman Swoop Talent Integration service to unify talent data sources

10 REASONS HUMANS ARE BAD AT HIRING (AND HOW AI CAN HELP)

It’s really easy today to get caught up in the headlines about AI being
biased. The truth is that artificial intelligence, in its current form, is ideally
poised to help minimize some of the very real biases that humans have that
limit our ability to hire effectively. Below are ten of the common biases that
affect human decision-making in every aspect of life, but in this scenario
the context is all around hiring and candidate selection:

1 Dunning-Kreuger effect: We think we are better than we really are at

making hiring choices, even though we’re just as bad as everyone else.
Hiring translation: ‘I am a really great interviewer. I almost always make
the right choice of candidates.’
2 Confirmation bias: We want to believe ideas that confirm what we think
to be true.
Hiring translation: ‘This candidate went to the best school, so they have
all the best answers to our interview questions.’
3 Self-serving bias: Others cause our problems, but we cause our own wins.
Hiring translation: ‘Jerry told me to hire Maria but she didn’t work out;
however, I hired Jamie and she was a great choice.’ In this example the
hiring manager may have overtly or subconsciously given Maria signals
that she wasn’t wanted, and she left because of them. Those same signals
weren’t given to Jamie because she was actually the desired candidate.
4 Self-handicapping: We could make the right choice but we have an
excuse in hand just in case.
108 ARTIFICIAL INTELLIGENCE FOR HR

Hiring translation: ‘I just knew Elvis wouldn’t work out for us. I wanted
to wait another month to pick but someone else made me choose a
candidate before I was ready. We were rushed to hire.’ This is all about
setting up an ‘out’ or an excuse in case it doesn’t work as expected.
5 Optimism/pessimism or headwinds/tailwinds: We overestimate likelihood
of good things and underestimate bad things; alternatively, we
overestimate the things that cause us to fail (headwinds) and
underestimate the tailwinds that set us up for success.
Hiring translation: ‘I hope we can hire a great crop of candidates this
college recruiting season, but we are missing a recruiter and just don’t
have the relationships we used to with candidates at key universities.’
In this example, the company overestimates the negative headwinds
and underestimates the longstanding relationship with the university,
the value of the right technology to support the process, and other
positive aspects that would have set them up for success.
6 Sunk cost: We cling to things we already have, even if they hold little to no
present value.
Hiring translation: ‘I know Bob isn’t really working out, but it’s just too much
hassle to hire someone else so let’s hold on to him for a while longer.’ I call
this the red button test. If you had a red button to replace that person
immediately, would you press it? If so, it’s time to replace them because
sunk cost isn’t a reason to keep something of little/no value.
7 Decline bias: A belief that things are worse or harder today than they
were in the past.
Hiring translation: ‘There is no good talent today. It’s harder to hire than
ever before. The war for talent.’ Etc.
8 Backfire effect: We become more convinced and double down after we
are challenged on a position.
Hiring translation: ‘I know she may not be the best candidate but she is
the best one we can find. We really just need to hire her as soon as
possible.’
9 Anchoring: The first thing we see influences how we evaluate everything
else.
Hiring translation: ‘That first candidate was really outgoing and a great
conversationalist. Lots of fun. None of the other candidates were really
as talkative as that first one…’
 TALENT ACQUISITION 109

10 Halo/horn effect: Happens when one attribute or trait overshadows others in

a positive (halo) or negative (horn) manner.
Hiring translation (halo): ‘That candidate did an amazing job on the
assessment. I think she would be great for any job we have available right
now.’
Hiring translation (horn): ‘Wow, that last interview was rough because the
guy was a smoker. I didn’t hear anything he said in response to the questions
because I couldn’t breathe.’

Admit it: we’re bad at selection. The data shows that the common ways we
interview and many of the methods companies use to rank candidates
(school attended, college grades or other demographic data) are highly
unreliable statistically. Translation: they are terrible as a gauge for whether
someone can do a job or not.
Instead, we should rely on more dependable data sources, such as job
samples (let someone try the job before they are hired to do the job),
assessments or structured/regimented interviews. If we can use these more
predictive types of data, we can make better hires and improve quality of
hire. And, interestingly enough, AI is positioned to do just that.
AI can help to mitigate or eliminate these types of biases if developed
with a diverse team and trained on an unbiased, comprehensive data
model. At a minimum, AI can help to limit the times when humans have to
make decisions in the absence of data, as the list of biases above proves
that we often can’t make accurate, rational decisions.
Reflect: Which of these biases is a problem for your team? For you? How can
you overcome them?

Sourcing tools and technologies

Sourcing candidates is one of the most data-heavy aspects of recruiting.
Consider the process a sourcer or recruiter often uses to find talent, the
various systems that data can be stored in, and the stages of communica-
tion that exist, and you’ll quickly understand why sourcing is the
low-hanging fruit of the talent acquisition world. A normal recruiter
spends an inordinate amount of time sourcing, the act of combing
through resumes and connecting with potential candidates. Having tools
on hand to sift through the massive amounts of data is highly valuable.
External sourcing is what recruiters typically mean when they talk
about recruiting. This isn’t about posting a job and waiting for
110 ARTIFICIAL INTELLIGENCE FOR HR

c­ andidates – it’s about going out and finding the right talent wherever
it might be. Let’s revisit the SourceCon Grandmaster Challenge. In this
instance, the contestants were attempting to find the right people for
fictional roles. As you recall, the humans won the competition, but the
AI-driven tool was not far behind. In the end, sourcing expert Randy
Bailey was crowned champion and another set of contestants tied for
second place. The algorithm produced by Brilent came in third place,
but it only required 3.2 seconds to deliver results while the humans
spent anywhere from four to twenty-five hours to research their submis-
sions.4 This story is a powerful testament to the value that both parties
bring to the conversation. In his review of the competition, sourcing
expert Jim Stroud mentioned that the humans that won or placed
highly in the competition used technology to narrow down the resumes,
but ultimately relied on human intuition or ‘their gut’ to make decisions
at the end of the process. In essence, there is a spectrum from totally
human to totally algorithmic, and the best value and performance lies
somewhere in the middle of that range. Let’s take a look at some of the
technologies that enable great recruiting performance today.
Examples of automated sourcing technologies run the gamut from
databases connected to smart search tools such as those offered by
HiringSolved to automated recruiting solutions like Envoy offered by
Entelo, with a variety of other ‘flavours’ of technology in between. The
core thread these various tools share is the blend of intuition and tech-
nology that enables performance greater than the sum of their parts.
Take the sourcing and search tools offered, for example. When
recruiters search a database for a specific set of software development
skills, they may or may not be aware of related skills and competen-
cies that would also be relevant to the search. For instance, software
development might be a top-level skill, but underneath that are
dozens of variations and specialties, from software testing to front-
end development. Because the system has processed resumes from
candidates with those skills, it knows when to suggest other, related
skills that might also help in a search. This may not hold a ton of
value for senior recruiters with a lot of experience – after all, they
probably already know the primary and secondary skills as well as
common recruiting channels from companies to universities to inter-
est groups. However, for more junior recruiters and non-technical
 TALENT ACQUISITION 111

HR professionals, those types of capabilities have the potential to

dramatically improve recruiting results, lower cycle times and
improve that all-important hiring manager relationship. While
searching candidate databases and using Boolean search tactics have
been mainstays of recruiting for some time, the real shift here is that
algorithms can infer skills matches in a way that humans simply
can’t. For clarity, Boolean search is the process of searching a d ­ atabase
using operators to narrow search results. For example, if you search
in a database for ‘software engineer’ but also use ‘-Java’ you can filter
out any software engineer resumes with Java in them. More complex
Boolean logic can give more accurate results.
A different option on the sourcing front is today’s slate of automated
sourcing and contact tools. In late 2017 Entelo unveiled its Envoy solu-
tion, designed to help busy recruiters reach out to more candidates in a
scalable, automated fashion. The common method for sourcing candi-
dates for a role is this: you post the job, you go to a database, you search
for candidates, you contact those promising candidates, you follow up
with any candidates via phone to move the ball forward and potentially
seal the deal. In reality, this is incredibly time consuming. For highly
technical roles I’ve filled in the past, the sourcing phase could take
anywhere from eight to thirty hours per position or more, depending on
just how specialized it was. Sourcing technologies attempt to shorten
that chain by automating a few of the more time-consuming aspects.
Using this software, recruiters post a position, then they take a well-
deserved break. In all seriousness, the technology picks up after the
search criteria are identified, automatically finding candidates in the
database that are likely matches for the skills and automatically reach-
ing out to them on your behalf. Any warm or responsive contacts are
then transitioned to the recruiter for a more ‘human’ touch. Ask any
recruiter and they’ll tell you the early parts of the sourcing process are
fairly straightforward, if you know how to search and make contact
with candidates. The actual high-value conversations don’t come until
later, but without having a solid funnel of candidates there will be no
opportunities to have those high-value interactions.
Another example of technology developing in the sourcing world is
what the team at Restless Bandit calls ‘talent rediscovery’. It is one of
several companies that offers this type of approach. The basic idea is in
112 ARTIFICIAL INTELLIGENCE FOR HR

rediscovering the valuable connections you already have in your own

database before paying to find candidates elsewhere through job
boards, job ads or other methods. Think for a moment about your
applicant tracking system, if you have one. How many resumes are in
there relative to the number of positions you’ve posted over time? Using
the estimate of 250 applications per posting from earlier in this chapter,
a company that has posted 100 jobs has more than 25,000 resumes in
its database. One problem with this is that of those 25,000 resumes,
maybe only 18,000 of them are unique, the rest are duplicate applica-
tions by the same person for different jobs. Yet another problem? Stale
data. Information on those resumes, if it’s left to stagnate over a few
years, may be virtually worthless in some cases. People change jobs.
They relocate. They change contact information. Without updated
information, those resumes lose their value over time. Therefore, these
systems rely on an algorithm that sits on top of the applicant tracking
system collecting data on hiring patterns and which candidates ulti-
mately succeed in the selection process. This feature can quickly analyse
a database of prior applicants and resurface the ones that are most
relevant to the current requisition. Logically, because these are candi-
dates that have already applied to the company at some point in the
past, it’s possible that they might be a ‘warmer’ audience than those
available through other typical sourcing channels and cold outreach.
Clients of these firms are able to rediscover candidates buried within
the applicant tracking system to fill new requisitions, saving time and
resources to locate new applicants.
Other opportunities on the sourcing front are more process-oriented,
though still incredibly valuable, as many firms have poorly defined and/
or poorly refined recruiting processes. For example, Beamery, a UK-based
recruitment candidate relationship management platform, also brings
value to the sourcing conversation. The company’s tools not only help
to automate the applicant ranking process, they also support interac-
tions with candidates. For instance, if recruiters fail to follow up with a
high-value applicant, the system can trigger an automated reminder to
get the applicant back on the recruiter’s radar. Because timing is critical
in recruiting, the technology can also suggest the most valuable time to
reach out and connect with candidates. Finally, recruiters can leverage
machine learning-powered data m ­ atching and automation systems to
 TALENT ACQUISITION 113

help them prioritize the candidates that are most engaged and are the
best fit for open and future roles.
Finally, no sourcing conversation would be complete without a
mention of LinkedIn, one of the default tools for virtually every
recruiting shop. At the end of the day, large players like LinkedIn are
sitting on a stockpile of data that can be leveraged along with AI to
create powerful tools and recommendations for employers looking to
hire the best talent. Additionally, they are often the first stop for
recruiters and sourcers looking to find the right candidates not just
because they have amassed large numbers of users, but also because
their search algorithms are fundamental in narrowing those pools of
users to the right candidates.
LinkedIn’s Recruiter tool uses a variety of machine learning to
power its recommendation engine. Beyond the typical resume
keyword search, the system can analyse a wide spectrum of informa-
tion to find and deliver the right candidates to a recruiter. Interestingly,
one newer signal that the company has brought to the product is
candidate receptivity. This has a double benefit of prioritizing people
that might be open to contact about jobs and deprioritizing those
that are not, improving the user experience within the platform. By
analysing a wide spectrum of behaviours, the system can surface indi-
viduals in searches that are more likely to be receptive to contact
from a recruiter. Some questions the algorithm considers in its recom-
mendations of candidate openness to recruiter contact are:
●●
How responsive has the individual been to recruiter contact in the
past?
●●
How responsive has the individual been to contact with various
types of recruiters, companies or industries?
●●
Is there any specific, recent job search activity?
●●
Are there any connection activities or trends apparent and what
might they signify?

This approach could help to improve hiring outcomes for employers.

Employer branding is focused on getting the company’s culture and
information out into the market so candidates can be aware of the
firm and its available jobs. It logically follows that candidates who
114 ARTIFICIAL INTELLIGENCE FOR HR

share, comment or interact with a company online are much more

likely to respond to contact from a recruiter. If algorithms push those
types of individuals higher in recruiter searches, that should lead to
better, more targeted contacts, conversations and hires. Additionally,
with news in 2017 that both Google and Microsoft have entered the
small and mid-size business market with hiring solutions, it’s not a
stretch to imagine some of those tools making use of the existing AI
infrastructure in these technology firms to make talent acquisition
more predictive and analytical. For instance, at the 2016 Microsoft
Ignite user conference, CEO Satya Nadella showed the audience the
integration of its Cortana AI with LinkedIn data to help someone
planning a meeting to better understand the attendees by visiting
their LinkedIn profiles. Additionally, the Google Jobs initiative to
bring an AI-powered search experience for candidates is another
example of a single step with far-reaching impacts.
While not an exhaustive list, these examples pretty clearly demon-
strate the type of value that AI can bring to sourcing. As talent
acquisition is one of the most labour-intensive elements in the entire
HR spectrum, professionals in this field stand to gain much from the
adoption of AI-based recruiting technologies. However, as we all
know, simply sourcing candidates isn’t enough. We need to be able to
match, screen and hire the right ones as well.

CASE STUDY
Automated job offers: Fantasy or reality?

If a recent graduate applies for a software engineering role in one of the company’s
competitive markets, the algorithm pushes the candidate through a series of short
assessments to understand coding skill levels and personal characteristics. If the
candidate scores above a certain threshold, a job offer is generated automatically
without any interview being required.
Note that in this entire process, not a single human interaction occurred, and
nobody from the company was involved in screening, in any form or fashion.
This sounds like a futuristic version of recruiting, but this example of an Amazon
hiring experiment was mentioned in HR Magazine in June 2016.5 For many
companies, this sounds like a somewhat frightening concept. After all, who
offers a job to a person they’ve never met or even spoken with? However, from
 TALENT ACQUISITION 115

an objective standpoint, this is an incredible solution to some of the problems

that have plagued the workplace for some time. For example:
●●
The offer was made at market rates with no regard for salary history.
Questions about salary history have been banned in several US states to
improve pay equality.6
●●
The offer was made regardless of the individual’s race, country of origin,
gender, etc. This effectively removes any bias from the hiring process.
●●
The person was reviewed objectively with a targeted assessment, ensuring
that core skills were apparent and that the applicant was not able to get the
job merely by performing well in the interview, socially or otherwise.

More practically, the experimental solution was put in place to solve a more
pressing factor: the time required to interview, evaluate and offer a job to a
candidate. In tight labour markets, time means lost opportunity in many cases,
and the fastest hiring process of all requires no human interaction.
While it’s still early, future advances on this type of approach could include a
few additional pieces of automation. For instance, what if the workflow looked
something like this upon receiving a candidate’s application:
●●
The job is historically difficult to fill.
●●
The position has had no qualified applicants to date.
●●
The candidate’s resume meets the basic skill requirements for the role.

The near-instantaneous analysis complete, the firm’s recruiting system supplies

a short assessment to the candidate to determine basic skills proficiency in the
specific role. Upon the satisfactory completion of the assessment with high
marks, the system automatically generates an offer letter to the candidate at the
current market rate.
While most of us are not ready for this level of sophisticated automation in
the hiring process, we can all agree that it diminishes the biases that plague
hiring decisions in virtually every company around the world. It’s a radical
approach, but then again it’s a radical problem as well.

Candidate matching
Similar to sourcing, matching is another time- and labour-intensive
activity for recruiters that can be best augmented by the right ­technology.
116 ARTIFICIAL INTELLIGENCE FOR HR

Matching is the process of pairing applicants with jobs and ranking

them based on their qualifications. In the past, most matching was
done by simple resume keyword extraction. The more keywords you
have in the resume, the higher you rate for job fit. While this isn’t a bad
place to start, there should be additional considerations in the process.
As someone that has had to review keyword-stuffed resumes in the
past for unqualified candidates, I believe there has to be a better way to
separate the wheat from the chaff.
This is also a critical point because as we get closer to an official
hiring decision, companies would in theory be less and less comfort-
able with allowing AI to have the final say in the conversation. After
all, at the end of the day the HR team and the hiring managers at the
firm are the ones on the hook from a compliance standpoint for
hiring decisions. Who’s going to put an algorithm on trial for dispa-
rate impact or discriminatory hiring? For that reason, the closer we
get to a hiring decision, the more humans will interact with the
systems to make a joint decision on the best candidates using the
intuition both systems can offer.
An example of this comes in HireVue’s video screening tools. Video
interviews are often thought of only in the live conversational context,
such as a live video chat between the candidate and the hiring team.
In truth, many employers prefer to use asynchronous video inter-
views to examine candidates in addition to and as a precursor to
those live interviews. Asynchronous videos are simply videos that
have been recorded at a previous time and that may be played at any
time. They are the opposite end of the spectrum from live video
conversations where both parties must be present at the same time
for discussion. HireVue’s Insights score can consider a range of candi-
date factors to rate their ‘performance’ in a recorded interview.
Considering things like tone, word use and even eye contact, systems
like HireVue’s allow employers to see candidate scores at a glance
before deciding which to follow up.
A similar video hiring tool, the mroads Pañña system, can score
candidate video performance, using machine learning to listen and
watch for aberrant behaviours. For instance, ways to cheat on a video
interview might be listening to cues from someone off-screen or
 TALENT ACQUISITION 117

a­ verting your eyes to read notes from a notepad. The mroads technol-
ogy can see and hear those types of actions and flag them, allowing
recruiters to double check on those video recordings to verify the
content. Another interesting capability of the system is a quick assess-
ment that candidates can take on screen or during the application
process. If candidates apply for a technical job, the system can flash a
quick coding quiz on the screen for the individual to complete, quickly
demonstrating their capability with the programming language associ-
ated with the position. The system automatically scores the candidates
on the questions upon submittal and ranks them in the dashboard so
recruiters know which candidates to prioritize in their outreach.
In the matching phase, it’s easy for humans to be swayed by a
number of factors, whether consciously or unconsciously. As mentioned
above, the resurgence of focus on diversity means technologies that
support these decisions with data and evidence instead of ‘gut feel’ are
going to become more and more valuable. A great example of this in
action is the set of tools offered by HiredScore. When individuals apply,
HiredScore captures their resume and application data. The system can
quickly analyse and understand which applicants match the job
requirements. Because this is a compliance-based system, it doesn’t
consider irrelevant factors and focuses on whether the person can actu-
ally meet the stated requirements of the job.
Simple concept, right? If a system can analyse inputs and match
resumes with job postings, we can select the right people for the job.
In reality, the problem that exists is that job postings and resumes tell
distinctly different stories. One tells the story of the employer, describ-
ing the history, cultural elements and strategic direction. The other
explains the career history of the applicant, demonstrating skills,
roles and other accomplishments. Analysis of the resume can pull out
key bits of information to match with the job description, but it’s not
as simple as it sounds. The thing that users want most from a system
like this is matching on an automated, ongoing basis with a thorough
look not just at making it simple and easy but also at the compliance
aspects. It shouldn’t fall back on the HR or recruiting leader to see if
bias or adverse impact is occurring – the system should be doing that
analysis all along.
118 ARTIFICIAL INTELLIGENCE FOR HR

Assessments
Another consideration on the matching front involves assessments.
Just as mroads and other video interviewing providers are incorpo-
rating assessments into the video screening process, other providers
are exploring automating this as part of the standard application
process. This include firms like Koru, Pymetrics and Fortay. The latter
offers a solution that aims to help companies scale their culture as
they scale the business, one of the most common questions or chal-
lenges for fast-growing employers. While ‘culture fit’ is a loaded term
in many HR circles, employers that understand some of the key
competencies and behaviours that drive business results can leverage
assessments to hire for those specific skills.
A quick word on ‘culture fit’, as some employers might get the
wrong idea about this. Some companies are shy about the ‘culture fit’
conversation, because they have seen it used in a negative manner by
hiring managers to block diverse candidates they didn’t want to hire.
However, good employers can still hire for key qualities that exist
regardless of racial, gender or other demographic lines. For instance,
employers that place a high emphasis on customer service are not
limited to any specific demographic group to find those qualities – the
fundamental value exists in a very broad population. When I talk
about culture fit, that’s the concept I’m referring to.
Back to the technology discussion: Fortay’s platform uses machine
learning technology to define a firm’s unique ‘cultural fingerprint’
based on ideal company values. Additionally, it examines and codifies
cultural workplace attributes of the company’s cultural champions,
those that embody the key aspects of the firm’s culture. This finger-
print is then used to assess relative cultural alignment in candidates
to ensure a successful match and increased quality of hire. When
clients use the system, they are able to improve the efficiency of the
hiring process while simultaneously improving the quality of the
candidate pipeline. For some clients, an unexpected benefit of lever-
aging Fortay was that it inherently enhanced the diversity of
candidates by minimizing the potential of bias in the screening
process for factors unrelated to the job or culture.
 TALENT ACQUISITION 119

CASE STUDY
Unilever uses AI for university hiring needs

When companies want to hire students and recent graduates, they always do the
same thing. The firm picks a university job fair, sends a recruiting team and
collects paper resumes from attendees at the event. But the process is costly and
the connection with results is somewhat tenuous. That’s part of the reason
Unilever decided to change its approach to campus recruiting by leveraging
AI-driven assessments and video interviews.7
The new screening method involved game-based assessments to examine
candidates on a range of measures, including tolerance for risk. Existing
employees took the assessments to create a benchmark to compare the
candidates against, a concept known as concurrent validity. Concurrent
validity helps employers to know if the scores candidates get on an
assessment would match well with the characteristics and qualities of existing
high performers on the team. Additionally, the video interviews were
AI-driven. Answers to the questions were analysed by an algorithm that
considered body language, words and phrases used in conversation and tone
of voice. The system then highlighted those candidates that were most
qualified for the next step of the process, signifying the first time a candidate
interacted with a recruiter at the firm.
The results have been incredibly impressive. After adopting the new
technological approach, Unilever was able to hire its most diverse class ever,
seeing a significant increase in minority applicants. Additionally, student hires
came from more universities than ever before. Traditionally the company
selected from just over 800 colleges, but the new crop of employees was hired
from more than three times that many educational institutions (2,600). One of
the most important metrics for any recruiting shop, time to fill, dropped from
four months to four weeks, saving an incredible amount of time both for
candidates and for the recruiters themselves.
Perhaps most importantly from a candidate experience perspective, a survey
of 25,000 candidates showed a score equating to 82 per cent satisfaction, far
beyond the typical satisfaction rates of most hiring processes for enterprise
organizations. This just goes to show that increasing adoption of automation
technologies doesn’t have to lead to lower satisfaction from candidates – it can
actually improve the results if approached strategically.
120 ARTIFICIAL INTELLIGENCE FOR HR

The role of data in recruiting

One thing is for certain after seeing just a small selection of the systems
available to solve today’s recruiting challenges: the future will be filled
with more data and inputs from these types of systems than ever
before, with many enterprise organizations juggling multiple systems
due to geographic, business unit or other specialized needs. But where
does that data reside? Is there a way to connect it in a usable manner
with organizational systems, or is it locked away as static informa-
tion? That’s the problem that systems like Swoop Talent are designed
to solve. Swoop Talent is an example of a system that connects dispa-
rate data sources to allow employers to have a single, unified view of
their talent and capabilities. While there are several systems in the
market that can bring various data sources together, what’s interesting
about Swoop Talent is the wide variety of use cases, limited only by
the imagination of the talent acquisition teams it supports.
As an example, one interesting use case for the system is to target
alumni, or former employees. Let’s say a software developer resigns, is
hired by the competition and begins gathering skills at another job.
One to two years from now, that person is more valuable than they
were when they left the role, correct? After all, the developer has most
likely been learning and improving the requisite skills throughout that
period of time. Eventually a position opens at the former employer
that the person wouldn’t have been a fit for. However, because the
employer is using Swoop Talent, it can see from the updated talent
profile (which pulls data from social sources like LinkedIn as well)
that the individual now has the required skills necessary to perform
the job. The recruiter can reach out to this person to start a discussion
about returning to the new role. With research showing that returning
employees are consistently a source of high-quality talent, this makes
perfect sense as a recruiting strategy to recapture valuable workers
even after they have left.8 Swoop Talent’s machine learning and predic-
tive tools help to bring more information to recruiters so they can
make better, well-informed decisions at the right time.
One final example while we’re on the topic of data: IBM’s new
offering in the talent acquisition space is IBM Watson Recruitment.
 TALENT ACQUISITION 121

The cognitive system increases recruiter efficiency by highlighting

priorities that need attention. One way the system can use machine
learning is to analyse applicant flow, previous recruiting funnel statis-
tics and other signals to indicate whether a job is on or behind
schedule from an applicant flow perspective. For instance, if it has
taken three months and fifty candidates on average to find a suitable
executive sales leader in the past, the system might flag the job if after
a month it only has four or five candidates, far shy of the target
necessary to stay on track with previous hiring cycles for the same
role. Additionally, the system provides insight into which candidates
are best qualified by automatically rating factors such as duration of
experience in relevant roles, size of previous employers, college degree
and major, and more. However, this algorithm doesn’t just march off
on its own without support. Recruiters can actually rate the quality
of the recommendations to help train the system to improve its
suggestions over time. For example, if college degrees aren’t relevant,
the recruiter can deprioritize that component while prioritizing expe-
rience at previous similar employers, which might be particularly
relevant for this example of a sales executive. IBM is known as one
of the frontrunners of leveraging AI in a variety of areas both inside
and outside human capital management. The firm’s Watson product
is connected with so many other vendors via partnerships or integra-
tions that it powers a surprising number of insights that we rely on in
the HR space.

Recruiting applications for natural language processing

While we have explored the idea of natural language processing to
some degree in a broader sense, it has some very interesting applica-
tions within the talent acquisition function. From augmented writing
to skills analysis, employers have much to gain. As mentioned previ-
ously, most companies are sitting on an incredible amount of
information that is just waiting to be tapped into. Tools that properly
analyse that data, whether structured or unstructured, can help to
illuminate insights and potential avenues to success.
122 ARTIFICIAL INTELLIGENCE FOR HR

One way employers can take advantage of NLP in recruiting is by

analysing and understanding how to write the best job ads. One of the
frontrunners in this space is Textio. Textio’s augmented writing system
is driven by examining more than 250 million job postings and their
hiring outcomes. The analytics engine then delivers contextual help to
recruiters by enabling them to craft better job ads. The system does
this by instantly analysing job posting text content and predicting
which words will have the best reception and impact on candidates.
For instance, if the score appears as a 60 out of 100, the system flags
words that might improve the score and performance of the ad, allow-
ing recruiters to quickly and easily modify the content on the fly. For
example, a single word may shift the perception of the ad to a more
masculine or feminine nature, impacting the types of candidates that
apply for the role. Because Textio knows the kinds of people that apply
for jobs with similar wording, the predictions are incredibly precise.
The value promise of this type of tool is faster, more diverse hires
of higher quality. This has become such a trusted asset for users that
one client now requires all descriptions to be published with a mini-
mum score of 90. Why so specific? It’s because Textio’s hiring data
shows that jobs with a score of 90 or above are filling 17 per cent
faster with up to 25 per cent more quality candidates in the pipeline.
Trusted by enterprise hiring teams at Johnson & Johnson, Atlassian
and other firms, the application gamifies the dull process of creating
and posting job ads.
In the last few years, we’ve seen other enterprise software vendors
bring these kinds of augmented writing tools to their own platforms.
As mentioned in an earlier chapter, when you can integrate these
capabilities natively instead of having to access a different program
or workflow to get your job done, you’re more likely to take advan-
tage of the technology. Technology adoption is always a challenging
prospect, but even more so when it requires using multiple applica-
tions to get the job done.
In addition, these kinds of tools hold incredible promise to support
better diversity and inclusion (D&I) outcomes. The way I’ve always
explained it to executives is this: there’s only so much wiggle room a
company has to improve diversity once the employees have been
 TALENT ACQUISITION 123

hired. We can transfer and promote diverse workers, but if you have
a homogeneous pool of candidates to work from at the front end of
the hiring funnel, there’s only so much advancement that can be
made. For that reason, it’s critical to make sure to keep diversity in
mind during the job requisition development process, one of the
earliest stages of the talent acquisition cycle.
One innovative example of AI that doesn’t fit neatly into these
other categories yet still leverages natural language processing is the
interview intelligence solution developed by BrightHire. The compa-
ny’s technology lays over the tool used to record the interview, such
as Zoom, and analyses the conversation. Think about your current
interview process: you know when it happened, who was there and
what kinds of written feedback were collected, but that’s pretty much
it. What if you had a system to look at how long your hiring manager
spoke versus how long the candidate spoke? That’s powerful to
understand whether men, women or minority candidates are getting
the same opportunity to share their expertise during the process. Or
what if it could tag specific parts of the conversation to allow quick
reviews of accomplishments across a set of candidates for compari-
son purposes? This tool does all that and more. I’m excited to see
where these types of systems evolve to create clarity and insight into
interviews because historically much of what happened during the
interaction was never captured or analysed in a meaningful way.
Another natural language processing use case is focused on skills.
Earlier in this chapter I explained the challenge of matching resumes
(candidate career history) with jobs (candidate skills). I talked about
this from a company perspective of matching candidates with job
openings, but there’s actually value in the process for candidates as
well. Imagine that you are looking for a job as a vice president of
human resources. You go to one of the job aggregator websites and
create an alert to email you with jobs that match your search criteria.
Better matching technology means you’ll get alerts only when jobs
exist that are relevant and interesting to you. However, many of the
tools in the market today will send regular emails even with irrele-
vant or incorrect job matches simply because one of the words in the
job advertisement matched your search criteria. Fixing this problem
124 ARTIFICIAL INTELLIGENCE FOR HR

helps to improve the candidate experience in a meaningful way. One

fundamental way it approaches this is by translating jobs into actual
skills, breaking down something as vague as a job title into the key
skills associated with the position.
What if you were going to buy a vehicle but didn’t know what the
fair market value was for the specific model and feature set you were
targeting? You know the basic details and ballpark price range, but
how do you narrow in on a price that is specific to the vehicle you are
looking for? More importantly, what if there was a system that gave
you transparency into the pricing not just for the specific model but
for each individual feature you consider to be a high priority? One of
the technologies in the market today can do this for jobs. Instead of
relying purely on job titles to price open roles, which has been the
common practice historically, the system allows users to see prevail-
ing compensation rates for specific skill sets and helps to highlight
and isolate the pricing impact of each. As an example, users can price
a job title and then start adding key skills one by one, viewing the
cumulative impact on the cost of the role. This transparency might
help users to realize that while they have always been listed as a
requirement, one or two skills might not be necessary if they are
pushing the cost of the role inordinately high. The practical implica-
tions on cost are staggering: one client using the system has
accumulated more than $6 million in savings by streamlining and
improving efficiencies in its approach to hiring. With an economy
that is increasingly dependent on contingent labour, it’s critical to
consider that piece of the puzzle when pricing a job opening.

Chatbots as a communication mechanism

Chatbots offer incredible value and time savings for recruiters, and
they also open up lines of communication with candidates in ways
that humans simply can’t without intense efforts of manual labour.
Many of these tools in a recruiting context help to screen candidates
at the top of the funnel, provide an interactive experience for
­applicants and keep recruiters updated in the back-end dashboard. In
one demo, I actually had fun experimenting with the SMS-based
 TALENT ACQUISITION 125

interaction as the algorithm helped to schedule me for my ‘interview’

with the firm. It made me pause, because I knew in the process that
this wasn’t a person, but the bot served as a proxy for interpersonal
contact in the recruiting process, which was good enough for me.
Even though I wasn’t applying for a real job and was just testing the
technology, I have to imagine that users that are actually looking for
work are going to enjoy the experience even more than I did.
The use case for bots in the talent acquisition process is straightfor-
ward: candidates interact with a chatbot instead of a person, allowing
them to get instant help and attention and freeing up recruiters to
focus on other activities. However, we’re starting to see more automa-
tion in the process, opening up a world of innovative ideas. One
example of unique usage of a recruiting bot was in automated screen-
ing of candidates for a complex role. In this instance, the ideal
candidate for the job required a fairly significant following on social
media to be qualified, so the bot asked the candidate about what social
media platform was their most active and what their username or
‘handle’ was. The algorithm would then quickly scan the user account’s
posting consistency and recency to ensure that they met the minimum
threshold for consideration. While the system also performed the
standard interactive chat conversation with candidates, this step added
value by automating a key part of the process, eliminating a time-
consuming activity for the recruiting team to follow up and review the
social media profiles of every potential candidate to ensure they had
the necessary qualifications to move forward in the process.
One more example comes from Anna Ott, VP People at HV
Capital.9 In her story, she told about some of the surprising findings
of adopting a chatbot for recruiting. For instance, candidates mostly
interact with the bot outside of normal working hours. This means
those applicants are getting attention when a recruiter would not
typically be working, which is intriguing. The company’s system,
called ‘hubbot’, also received a wide range of questions beyond what
was initially expected. Bots are programmed to respond to specific
questions, which meant there were other queries outside its normal
parameters it could not answer. If the question that comes in is not
able to be matched to an answer in the database, hubbot sends the
query directly to Ott, who answers it and then programs the response
126 ARTIFICIAL INTELLIGENCE FOR HR

into hubbot for future enquiries. The breakdown of questions from

candidates probably wouldn’t surprise any long-time recruiters, but
it’s intriguing nonetheless:
●●
salary 27 per cent
●●
job specifics 24 per cent
●●
company information 20 per cent
●●
recruiting process 11 per cent
●●
work culture and office environment 9 per cent
●●
chatbot-related questions 7 per cent
●●
general questions 2 per cent

Ott’s analysis of using the bot concluded with this valuable insight:
employers can (and should) focus on being human while leveraging the
advantages of bots to meet their hiring goals. In each of these examples
of how AI is being used today within the recruiting sphere, it’s clear
that there is value in having these systems to help us engage with candi-
dates, helping recruiters to save time for more valuable, timely tasks.
This can include anything from recruiting high priority roles to having
conversations with key candidates in the pipeline, neither of which
would be suitably offloaded to a bot, algorithm or machine.

CASE STUDY
Southwest Airlines saves 90,000+ hours with recruiting chatbots 10

People crave personal connections when searching for a job. A quick phone call,
a handwritten note or a thoughtful text can make the difference between an
outstanding talent experience and a transactional one. The problem? Delivering
this level of personalization consistently and at scale isn’t sustainable for most
talent teams. But help in the form of automation and AI is making a huge
difference in conversational recruiting for the proactive companies embracing it.
In particular, chatbots (those virtual assistants popping up on careers sites
everywhere) are proving they’re worth their weight in gold.
Case in point: Southwest Airlines. Since implementing a recruiting chatbot
from Phenom, an award-winning talent experience management platform, to
enhance their candidate experience, the company reports over one million
interactions with it. This translates to as many as 92,000 hours saved for
Southwest recruiters, their team members say.
 TALENT ACQUISITION 127

Why chatbots? Chatbots are among the most visible applications of AI

transforming the HR industry - but there are a lot of ways to boost the candidate
experience. The biggest reasons Southwest wanted a chatbot on their career site
are also the technology’s greatest benefits:
●●
Lead capture. ‘We knew that every person visiting our site was a potential
candidate and employee,’ Sarah Steinmann, a Southwest Airlines Talent
Acquisition Specialist, said. Even if a right-fit position wasn’t on the site at
that moment, the TA team wanted to have a way to follow up when a
potential job match became available. By capturing valuable candidate
information and inviting them to join the Southwest Talent Community, the
chatbot can help meet that goal. The chatbot collects a candidate’s name,
contact info, job title, skills, years of experience, job category and location.
●●
Site personalization and job recommendations. When it comes to the
candidate experience, personalization is everything. Southwest knew that
serving up relevant content is key to keeping job seekers engaged and on the
career site longer. Once the chatbot gathers candidate information, it
generates tailored job matches on the spot, saving candidates the hassle of
sifting through superfluous opportunities - a priceless and much appreciated
perk for any job seeker.
●●
24/7 resource for FAQs. Southwest’s TA team saw the chatbot’s 24-hour
ability to answer questions as a major value-add. ‘We had thousands of
candidates visiting our career site without a really easy, clear way to find
answers to their questions,’ Steinmann said. ‘We wanted to explore if a
chatbot could help candidates get the answers they needed to know if a job
at Southwest could be a great fit.’

Delivering a recruiting chatbot requires three components: training,

collaboration and testing. Like other AI-driven technologies, the set up and
‘training’ of a chatbot is vital to optimizing the ultimate experience it delivers to
give candidates – and the value of its data for recruiters. ‘This is a really manual
process,’ Southwest Airlines talent acquisition team member Shannon O’Bryant
admitted, ‘but it was well worth the time and effort to ensure that the chatbot
was set up for success.’
The team at Southwest leveraged three key steps:

1 Building an FAQ workbook. Like many projects, this one started with a
spreadsheet, O’Bryant said. Recruiters and coordinators helped generate
questions for an FAQ workbook, drawing on sources like LinkedIn, interviews,
and back-and-forth emails with job candidates. The recruitment marketing
team then categorized the questions and wrote answers.
128 ARTIFICIAL INTELLIGENCE FOR HR

2 Collaborating between teams and departments. The TA team made sure their
departmental customers reviewed the accuracy of answers loaded into the
chatbot. This was a major key to success, according to O’Bryant. Once the
workbook was loaded into the bot, they launched a two-week testing phase.
3 Creating a dedicated testing team. Recruiters were assigned questions to ask
the chatbot and rate the answer with a thumbs-up or a thumbs-down. This
feedback helped train the bot on performance on the front end, while the TA
team matched answers to questions the bot couldn’t answer on the back end.

After all of that work comes the fun part: measuring success. Through the
process, Southwest learned that a properly tuned chatbot provides value to both
candidates and recruiters, Steinmann said. And their top wins illustrate the
mutual benefit across both candidate and recruiter experiences:
●●
A source of truth for FAQs. Southwest recruiters had been spending an
exorbitant amount of time answering questions from candidates about flight
attendant jobs - their most coveted and competitive position. The high
volume of questions still left certain asks unanswered, spurring flight
attendant job seekers to create their own groups on Facebook and career
forums dedicated to answering related job application questions. The TA team
was thrilled that the chatbot could ‘create a source of truth’ directly on their
career site to answer questions in real time for candidates, O’Bryant revealed.
●●
Time savings. As of their presentation at the Phenom user conference
IAMHR, the chatbot had seen more than 1,115,000 interactions. ‘If we
estimate that recruiters would spend one to five minutes answering each
question, it means the chatbot has saved us between 18,000 and 92,000
hours,’ Steinmann remarked. ‘Having a measurable way to give time back to
recruiters proved to be a huge value-add.’
●●
Direct applies. Approximately 28 per cent of career site visitors who interacted
with the bot clicked on a job suggested by the bot. Of those who viewed the job,
57 per cent went on to apply. ‘Our chatbot didn’t just have a random effect on
our hiring processes, but it really directly impacted the candidates we saw and
the future employees of Southwest Airlines,’ Steinmann said.

By being able to better engage and drive applies with highly motivated
candidates, there’s no denying the value of recruiting chatbots - and their ability
to empower talent teams with something nobody can ever seem to get enough
of: time for the moments that matter.
 TALENT ACQUISITION 129

Chatbot metrics to consider

For a long time, one of our priorities has been pushing for better, more
mature measurement within talent acquisition. The value of knowing
your baseline (or that of your industry) can help to guide decisions,
influence strategy and impact investment. Chatbots make data easier
to gather. Instead of relying on a recruiter to capture their time cost,
or other metrics in a report, chatbot data is gathered 24/7 and avail-
able for review at any time. That said, the metrics probably aren’t the
same as those you might be using today. Below is a set of potential
metrics that can help to evaluate the success and value delivered by
your chatbot in support of your talent acquisition activities.
●●
Total contacts/conversions: One key value point of a chatbot is
connecting with candidates who might otherwise leave without
applying. Many employers see bounce rates fall significantly by
leveraging chatbots to converse with passive job seekers on the
careers page. How many conversations lead to actual applications?
Compare this data with your careers page bounce rate.
●●
Session length: How long does the average user spend interacting
with your bot? How long does it take to get to the end of the process?
Is the length causing a drop-off in applications? We know that long
applications can cause candidate drop-off, so keep that in mind.
●●
Peak activity: When are candidates most frequently interacting
with your bot? Is it during normal work hours, or Sunday nights
at 10pm when your recruiters are in bed asleep? How can you use
this to inform conversations with candidates or deliver more
meaningful information via the bot?
●●
Conversation topics: Look for common topics and use those as
idea-starters for how to proactively address candidate questions or
concerns. Your results may vary, but knowing the types of topics
people ask the bot about can help to determine what your candidates
want to know via phone and face-to-face interviews as well.
●●
Handoffs: How often does your bot get stuck on a question or
comment and refer it to a human for help? Are these common
questions you should have programmed, or random ones you
couldn’t have foreseen?
130 ARTIFICIAL INTELLIGENCE FOR HR

●●
Time saved: How much time is being saved by leveraging bots?
Perhaps most importantly, how can you use that time to improve
the recruiting function, improve business alignment and create a
more engaging candidate experience?

KEY POINTS
●●
Due to the volume of applications relative to job openings, talent
acquisition offers a wide variety of opportunities for automation through
AI technologies, from candidate screening to matching and more.
●●
One targeted application of machine learning technologies in recruiting
is in reducing bias and improving diversity. In order to solve the bigger
issues with diversity in the workplace, we have to start with recruiting
and move forward from there.
●●
Recruiting technology applications are incredibly varied and plentiful,
serving niche use cases in almost any conceivable area. At the same
time, larger players in the market are weaving in artificial intelligence to
improve their own offerings to capture more market share.

Notes
1 J Sullivan. Why you can’t get a job… Recruiting explained by the numbers,
ERE Recruiting Intelligence, 20 May 2013. www.ere.net/why-you-cant-get-a-
job-recruiting-explained-by-the-numbers/ (archived at https://perma.cc/
7YC9-TPZ3)
2 Talent Board. 2017 Talent Board North American Candidate Experience
Research Report, 2017. https://icoportal.s3.amazonaws.com/files/
NHJgLCVQ/2017-Talent-Board-NAM-Research-Report-FINAL-180130.pdf
(archived at https://perma.cc/DS24-KGJ9)
3 M Bertrand and S Mullainathan. Are Emily and Greg more employable than
Lakisha and Jamal? A field experiment on labor market discrimination,
American Economic Review, September 2004, 94 (4), 991–1013. www.
aeaweb.org/articles?id=10.1257/0002828042002561 (archived at https://
perma.cc/CNY6-WZ2Z)
 TALENT ACQUISITION 131

4 J Stroud. Human vs machine: Who sources best? The results of the 2017
grandmaster competition, SourceCon, 21 March 2017. www.sourcecon.com/
human-vs-machine-who-sources-best-the-results-of-the-2017-grandmaster-
competition/ (archived at https://perma.cc/6ZDZ-WBWZ)
5 T Lee. 12 recruiting tips from talent acquisition leaders, SHRM, 1 July 2016.
www.shrm.org/hr-today/news/hr-magazine/0716/Pages/12-recruiting-tips-
from-talent-acquisition-leaders.aspx (archived at https://perma.cc/VC4J-
MPDQ)
6 HR Dive. Salary history bans: A running list of states and localities that have
outlawed pay history questions, 30 July 2021. www.hrdive.com/news/
salary-history-ban-states-list/516662/ (archived at https://perma.cc/P3ZT-
9UTN)
7 R Feloni. Consumer-goods giant Unilever has been hiring employees using
brain games and artificial intelligence – and it’s a huge success, Business
Insider, 29 June 2017. www.businessinsider.com/unilever-artificial-intelligence-
hiring-process-2017-6 (archived at https://perma.cc/F9R7-5RQX)
8 J Sullivan. Want a great source of quality hires? You need a ‘boomerang’
program, TLNT, 21 May 2014. www.tlnt.com/want-a-great-source-of-quality-
hires-you-need-a-boomerang-program/ (archived at https://perma.cc/UZ2Z-
ZPCA)
9 A Ott. A year with our recruiting chatbot, LinkedIn, 9 October 2017. www.
linkedin.com/pulse/year-our-recruiting-chatbot-anna-ott/ (archived at https://
perma.cc/ZY4X-MPES)
10 https://www.phenom.com/blog/the-value-of-one-million-chatbot-interactions-
with-southwest-airlines (archived at https://perma.cc/JBT2-246S)
132

THIS PAGE IS INTENTIONALLY LEFT BLANK

 133

06

Learning and development

Angel arrives at his tutoring session ready to dive in. The seven-year-
old has been making progress on his subtraction with regrouping and
is excited to see how well he can perform on the day’s assignments.
He spends some time initially in a small group setting, learning from
a tutor with several other students near his level. After that, Angel
moves on to computer games, honing his skills with practice. At the
end of the day’s session, he completes a short assessment to measure
his progress before going home.
That same day, Maria is also scheduled to attend the tutoring
session to help with her maths skills. She starts off with computer
games and then moves to one-on-one time with an instructor. At the
end of the day, she completes her own assessment to measure her
progress.
Both of these students are in the same programme, but both of
them receive different blends of instruction methods. Why? Because
it’s how they each learn best. This activity is part of the School of One,
a project based in New York City and designed to help tailor teaching
methods to each individual student. Unlike a normal classroom where
every child receives more or less the same type of instruction, this
programme gives students a highly personalized experience. The
concept of individualized instruction is simple, even if the algorithm
powering the learning delivery recommendations is fairly complex.
Students receive a variety of instruction via some of the methods
outlined above, and the regular assessments help to c­larify which
134 ARTIFICIAL INTELLIGENCE FOR HR

types of instruction are most suitable for each individual student. If

Angel improves significantly on a day he has small group, the algo-
rithm can recommend additional small group time in the next session
to see if improvements continue to trend. Alternatively, if Maria is
really doing well after computer games, the algorithm can pare back
other types of delivery like one-on-one tutoring by weighting instruc-
tional delivery more heavily towards the methods that her brain
prefers most. Excitingly, this gives teachers near real-time data on
student performance and learning to help guide them towards mastery
of skills on a repeatable basis.
This level of customization can be challenging to deliver, but the
exciting outcome is that each individual participant is given the type
of learning modality that best fits their need. What’s interesting to
consider is a more complex variant where Angel learns subtraction
with regrouping best by computer game but needs one-on-one
instruction to deal with fractions and small group time to focus on
multiplication. In reality, different topics may lend themselves to
different learning methods in the same way that different learners
may have their own unique set of preferences.
What does this have to do with today’s business environment? As
you may have guessed, the practical application of this type of adap-
tive learning has incredible implications not just for learning as a
discipline but for the outcomes of learning, such as engagement and/
or performance. Plus, it’s entirely enabled by a smart algorithm that
continues refining as more data is introduced to the system on a regu-
lar basis. We as business leaders might be able to look for trends in
the data and offer some customizations for certain learners, but
there’s no way we could do it individually for a large population. In
one interview, Jesse Jackson pointed out that he wants to use technol-
ogy to be more predictive in how learning paths and learning
opportunities are offered to employees. This isn’t based purely on
what the employee likes or enjoys, but on the type of learning assets
and experiences that will lead to the best performance and long-term
career success. Jackson is the Chief Learning Officer for consumer
and community banking at JPMorgan Chase, and he sees technology
as a critical tool to support this transition from more transactional
 LEARNING AND DEVELOPMENT 135

learning to high-impact learning that drives business results.1 Jackson

calls this the ‘business of learning’, and it is based on identifying the
behaviours that lead to performance and developing tools to opti-
mize performance to a higher level.
Let’s take the discussion in a more ‘meta’ direction for a moment,
because I need to touch on an area that’s applicable across the busi-
ness. Learning is at the heart of the adoption of artificial intelligence,
if for nothing more than the need for upskilling and reskilling the
workforce to adapt to the changing technology landscape.2 Wal-Mart,
the largest employer in the world, has invested more than $2.7 billion
to train its workforce on how to use newer technology offerings in
addition to reskilling workers to better serve customers.3 While most
of the headlines around the adoption of AI revolve around fear (AI
will take all our jobs, robots are coming, and the like), the truth is we
don’t yet know the full extent of the impact of AI adoption on the
greater workforce. What we do know is that as technologies continue
to come into play as I have highlighted throughout this book, L&D
and talent development practitioners will have their hands full
migrating the skills of workers to more broadly applicable capabili-
ties. As Galagan points out in her writing, some areas might include
helping learners manage their attention in an increasingly noisy digi-
tal workplace, or helping steer workers towards future careers that
are less susceptible to disruption by AI technologies. For more on this
specific topic, in Chapter 9 I examine some of the key skills that
talent leaders and the broader workforce should be cultivating in
order to ‘future proof’ themselves in the age of AI and automation.
Another area I would expect to grow is helping workers have more
of an awareness of how artificial intelligence can augment their roles
and how to best blend the human and machine components. How
can we help our workforce to work alongside algorithms to improve
performance, efficiency and results? I talk briefly about this concept
of performance support technology later in this chapter. In essence, I
see it as a highly tailored version of the content of this book delivered
on a highly practical level to jobs and professions across the entire
workforce. From the top executives to the early-career interns just
getting into the business, learning professionals have the opportunity
136 ARTIFICIAL INTELLIGENCE FOR HR

to help each group really understand AI technologies and how they

might get the most out of leveraging them for their jobs. More impor-
tantly, it may even help to elevate the performance of those workers
at the lower end of the performance scale. In an interview with a
leader of a recruiting technology firm, the CEO explained that he sees
artificial intelligence and machine learning as a way to categorize and
replicate the performance practices of the best workers, helping to
distribute those competencies, behaviours and skills to the rest of the
workforce. By definition, that would mean that your lower perform-
ers start to improve because they have the opportunity to mirror
those practices used by their higher-performing peers. Because L&D’s
reach and impact are stratified across the organization, the function
can have a greater impact on the business and its ability to react in an
agile manner to this monumental shift.
Having a smart technology looking at the problem beyond the
walls of the learning department is important, because learning ulti-
mately isn’t just a tool – it’s a fundamental part of work itself. While
formal content may come from the learning team, informal learning
takes place every day as a part of the natural rhythm of work.
Ultimately, learning in a corporate context isn’t just about cranking
out more content or more courses. That kind of quantity might have
been the answer to a problem in the past (unlikely, but I’m giving us
the benefit of the doubt), but today’s learning leaders are more and
more focused on quality content and learning assets as a way to
enable better business performance. It’s not just to ‘check the box’ on
delivering a set of training courses, it’s an opportunity to help each
individual drive the organization’s goals. As one learning executive at
McDonald’s, Kevin M Yates, is fond of saying, he’s had business lead-
ers come to him in the past with a problem and a request for ‘two live
classes and a side of webinars’ to resolve the issue. In reality that may
not address the problem, fix the problem or even be related to the
problem. In those instances a deep level of curiosity is necessary to
dig into the root of the issue and potential options to solve it, whether
that includes training or not.
 LEARNING AND DEVELOPMENT 137

An interview with Chris Ponder, Head of Global HR Compliance

at Moneygram International, highlighted a similar sentiment. All too
often, training is seen as the solution to a problem when it may just
be treating a symptom of a larger issue. Managers constantly asking
questions about a process? Offering them training may fix the issue,
but it also might lead to poor decisions because the process is ulti-
mately broken. An expert in performance improvement, Ponder
believes that talent and learning leaders have to get better at practices
like the ‘five whys’ concept where the person trying to solve the prob-
lem continues to ask ‘why’ until they reach the true root issue. For
instance: We have a problem with our learning content. Why? Because
it’s not improving our employees’ performance. Why? Because learn-
ing content is requested by managers on an ad hoc basis. Why?
Because we don’t have a strong governance procedure to prioritize
the right types of learning to meet our strategic goals. Aha! If we only
stop at the first question where we have trouble with our learning
content, we may try to change the modality, increase the amount of
content, or something else. But by looking deeper we can ascertain
the true cause of the problem and work to solve the real issue, not the
surface-level problem that’s easy to detect.
These are questions that humans are designed to solve, but that
doesn’t mean there is no place for AI technology in the learning realm.
On the contrary, the applications are wide and varied, supporting
everything from informal learning and collaboration to formal learn-
ing and content development. In terms of breaking down learning
into specific areas of focus, that’s a fairly logical way to do it. The
widely known 70–20–10 model, a staple of workplace learning
theory, estimates that 10 per cent of learning at work happens from
formal learning, 20 per cent from social experiences and collabora-
tion, and a whopping 70 per cent from experiential activities and the
flow of work itself. For that reason, we’ll break down our analysis of
AI applications and opportunities into formal and informal (social
and experiential) applications. But first, let’s consider what is perhaps
the most high-impact area of AI application within the realms of
L&D: the skills gap.
138 ARTIFICIAL INTELLIGENCE FOR HR

Closing the skills gap

In early 2021 our team surveyed 1,000 learners across North America
and Europe to learn about their preferences for learning and career
development. The findings were intriguing because they had the
opportunity to tell us how they wanted to learn new skills to close
gaps. Interestingly, it aligns with the 70–20–10 model mentioned
above. In today’s talent-driven employment environment, it’s critical
for employers to incorporate how people want to learn and grow
into their overarching development strategies. Table 6.1 shows the
learners’ responses by priority.
The skills gap is a function of the workforce not having the specific
capabilities the organization needs at a macro level. Therefore, a major
area where AI can contribute to the learning function is in helping to
minimize or even close the skills gap. Virtually every company would
agree that they have some sort of skills gap in their organization. In
one research study, my team found that the number one driver of
learning content demand for employers was to close skills gaps.4 For
purposes of a definition, the skills gap is the difference between the
skills mix that workers currently have and the skills mix the employer
needs to accomplish its goals. That gap can be costly. A CareerBuilder
study showed that the average cost of the skills gap to employers was
around $800,000 annually.5 The study looks at this from a recruiting
perspective, because that’s CareerBuilder’s target market, but the
concept isn’t much different from what we would apply from a

TABLE 6.1 Skill development methods, ranked

Priority Learning/talent professionals Learners/employees

#1 On the job (hands on) experiences On the job (hands on) experiences

#2 Coaching and mentoring Coaching and mentoring

#3 Company-provided training Company-provided training

#4 Internal mobility opportunities Internal mobility opportunities

#5 College courses College courses

#6 Boot camps/certifications Boot camps/certifications

 LEARNING AND DEVELOPMENT 139

l­earning context. In the study, the $800,000 price tag includes every-
thing from productivity loss and lower morale to higher turnover and
revenue loss from the inability to hire the right people with the right
skills to meet the needs of the employer.
How many of those companies could minimize or even solve their
skills gap problems by leveraging training and development instead
of trying to hire those individuals off the street? According to Josh
Bersin’s keynote presentation at the 2018 SkillSoft Perspectives event,
it costs employers six times as much to hire for the skills they need as
it does to develop them through training. If we combine this with
data from the Wharton School, we see that external hires not only
cost more money, they also perform worse on average in their first
two years on the job.6 Between those two data points, it’s astonishing
that any employer prioritizes hiring over training to the degree that
they do.
Now, here’s where it gets really interesting. If we bring artificial
intelligence into the conversation, imagine if the system was able to
look at the job openings posted on the firm’s applicant tracking
system, highlight the critical skills gaps the firm is trying to hire for
and start surfacing training for workers in those departments
where the openings occurred or even across the enterprise. While
the company is trying to find the right people through its recruiting
efforts, the firm is also training its workers to flex their skills and
grow their competencies to fill those gaps in the short term or even
in the longer term, if the positions are particularly difficult to
recruit for. Another version of this could be searching the employee
population for those with similar skills and abilities to the highest-
performing individual that left the job and caused the vacancy to
occur. For instance, it may be that someone working in a financial
institution is more likely to be high-performing if they have previ-
ous experience in the insurance industry or within a technology
firm, and the system can position those individuals as potential
internal transfers that can be developed into the roles through
internal mobility.
140 ARTIFICIAL INTELLIGENCE FOR HR

CASE STUDY
IHG combines assessments to quantify skills demand for diverse brands

IHG, a global hotel and leisure brand with nearly 400,000 employees, leverages
automated assessments during the hiring process to help weed out candidates
that do not fit the personality characteristics of its multitude of diverse brands.7
It also brings these tools to bear on reducing individual skills gaps.
Forget, for a moment, the wide spectrum of personality and skilled needs
within a single hotel. Front desk clerks, bell hops, cleaning staff, kitchen
staff – set those groups aside for a moment. Now consider the fact that IHG has
more than a dozen different hotel brands under its corporate umbrella. What
constitutes performance success for someone in a Holiday Inn may be different
for someone at a Kimpton Hotel, for instance. So how do you solve that?
The assessments IHG uses target the key personality traits that fit the needs
of the specific hotel chain’s brand, and then the company looks at skill
availability as a measure of candidate quality. In the perfect world, candidates
will be a culture or personality fit and a skills fit. But in reality companies face
decisions every day that have to do with picking a highly qualified person who
doesn’t fit the culture versus picking a person lacking in some key skills who
really fits the culture well. In this example, maybe there are two candidates in
the running for a specific opening. One of them has ten years of experience
managing a hotel but doesn’t quite fit the personality and cultural characteristics
of the hotel brand the individual has applied to work for. The other person only
has two years of hotel management experience but it was in a substantially
similar hotel and has the right attitude and values. Both are qualified in their
own way, so how do we select the right person?
IHG has committed itself to filtering out those poor fit candidates and
dedicating its learning and development resources to training for the right skills
to get other candidates into shape. Again, while there may be a core type of
training that each brand needs to establish a base level of skills in candidates,
each of the thirteen different brands has its own performance standards and
expectations for workers. This complex need is one that is best suited to
applications of artificial intelligence.

Formal learning applications for AI

Formal learning is what most of our time, effort and resources go to
as learning leaders. After all, formal learning encompasses e-learning,
 LEARNING AND DEVELOPMENT 141

instructor-led content and more. These formal activities also include

content. Lots and lots of content. For perspective, PwC has approx-
imately one million pieces of learning content in its library. The
sheer volume of content the company has generated is hard to wrap
my mind around, but times are changing. In the past it was enough
for learning teams to produce or purchase volumes of content and
present it to the learner. Mission accomplished. Today, though,
more than ever, learners are circumventing the l­earning manage-
ment system in favour of Google or other preferred sources of
information. Additionally, our research at Lighthouse shows that
low-performing companies are seven times more likely to say their
learning content doesn’t engage ­learners or is highly transactional.
Great content helps to create an engaging experience for learners.
In an interview with one global L&D leader, he explained that this
increasing trend of focusing on ‘experiences’ was borderline rubbish.
When pressed, he explained that no learning executive would get
budget for technology or resources merely to make learning more
pretty or engaging for the sake of it. However, if the outcome of
creating a more engaging experience was that learners retained
more information, were more likely to change behaviours, or more
likely to improve performance, then businesses would be much
more willing to invest in the concept of learning experience design.
This jives with another finding in that same research, showing that
high-performing companies are 50 per cent more likely to say that
engaging learning content helps to enable consumption of critical
learning resources. Simply put – learning experiences matter.
One way to measure learning experiences is by examining adoption
rates of learning content. Voluntary consumption would indicate more
engaging content and experiences, but as you can see in Figure 6.1,
many companies do not use this practice regularly. High-performing
firms are about ten times more likely to look at adoption rates to
­measure the impact of learning content, and they are more likely to
measure every facet of learning than low performers.8
142 ARTIFICIAL INTELLIGENCE FOR HR

FIGURE 6.1 Creating engaging learning experiences

Relevance for learners

and business needs
Better performance
by closing gaps

Adoption rates

Completion/compliance

Satisfaction surveys

Don’t measure

0% 10% 20% 30% 40% 50% 60% 70%

High-performing Low-performing

SOURCE Lighthouse Research & Advisory, 2017 Learning Content Strategy Study (n = 152)

One type of learning content that employers have and will continue
to rely on is instructor-led training (ILT). A variety of research points
to instructor-led training as the most commonly used and most effec-
tive method of instructing workers, so we’ll start there. Instructor-led
training is the most familiar type of learning process for us. We’ve
received years of schooling, and this is often the default method
employers use for sharing information and training throughout the
organization. While artificial intelligence probably won’t be taking
the place of the trainer in the front of the room any time soon, there
are roles that AI can play that will improve the efficiency of ILT.
One firm, Training Orchestra, offers a technology that supports
employers’ need to manage the various disparate resources connected to
offering live training. Think about the logistics of planning a training
class: you need an instructor, a space, a way to track registration, materi-
als, presentation tools, attendees, and more. It’s no surprise then that
most employers don’t make efficient use of those resources. Your top
trainers and most sought-after training spaces might be leveraged heavily,
but there’s almost always room for optimization and improvement.
Training Orchestra manages more than $5 billion in training investments
across its client base, helping customers to get the most out of their
resources by helping with scheduling, planning and resource allocation.
 LEARNING AND DEVELOPMENT 143

AI embedded in this type of technology would offer some incredi-

ble benefits. For instance, classes with open slots for attendees might
be flagged and prioritized so trainers and managers know where to
send available workers. Or maybe the AI sees that there’s a consistent
no-show rate of 10 per cent for one of the trainers, so it overbooks
classes by 5–10 per cent to help avoid lost opportunities. Another
option? The AI realizes a class is not going to be full so it can move it
to a smaller room to save the larger space for other activities, auto-
matically notifying participants about the change. Again, ILT is a
major portion of the average learning budget. According to Stephane
Pineau, CEO of Training Orchestra, formal training makes up 65 per
cent of the average learning budget but employers don’t do a great job
of making sure they get the most out of their systems and processes.9
Companies that partner with the firm see an average of 20 per cent
optimization the first year and incremental changes after that. While
your next course probably won’t be taught by a robotic instructor,
there is clearly a valid opportunity for AI to support the process.
For many years, the default piece of technology companies buy to
support their L&D needs is a learning management system. The
purpose of the tool is to deliver content and track completion rates,
but not much else, which is why there has been an explosion in
learning experience platforms in the market in recent years. In a
discussion with the visionary CEO of Docebo, a learning technology
firm headquartered in Toronto, we discussed the concept of a fully
automated learning system. Claudio Erba started Docebo in 2005
and since then has led the market with a variety of innovations, from
incorporating informal learning components into the system to
looking at AI elements that can radically change the expectations for
administrators and learners. His belief is that we should start with
the highly transactional processes administrators have to manage,
such as learner enrolments, for automation. Unlike some that want
to focus on the automation from the learner end, Erba wants to
focus hard on automating and augmenting the capabilities of L&D
professionals to make their lives easier and enable them to focus on
more strategic activities.
144 ARTIFICIAL INTELLIGENCE FOR HR

CASE STUDY
AI enables teaching at scale

Teaching is greatly enhanced by creating personal connections with students.

But how can instructors of college courses with hundreds of students create a
more personalized experience? For instance, in just one online Georgia Tech
class, students interact with teachers, teaching assistants and each other
through online forums. However, the volume of interaction is staggering. Just
one virtual class in the computer science department can have up to 350
students per semester, and discussion forums are incredibly active. To help
professors manage the workload, the university has authorized the use of
teaching assistants to support the demand. One of those assistants who recently
started working with students was Jill Watson. Jill was conscientious, diligent and
always courteous in her responses to student requests. In fact, she was such a
great help that the professor nominated her for teaching assistant of the year
out of appreciation for her efforts. He said he was very interested in seeing how
the university handled Jill’s nomination, since this was the first time a
non-human was nominated in the history of the programme. Yes, Jill was a bot!
And, as you may have guessed, she was incredibly effective at her job. But it
didn’t start out that way. The story below tells of her conception, issues and
triumphant finish.10
During one semester, those 350 students per class post an average of 10,000
messages. That’s essentially the same as receiving 100 emails a day for 100
days, and the university estimates it would require employing an instructor full
time for a year to respond to all of those messages individually. This caused the
course instructor, Ashok Goel, to wonder if AI could be used to answer the most
frequently asked questions. Since new student cohorts ask virtually the same
questions over and over again, the hope was an AI assistant might help to
alleviate the strain of responding to thousands of messages and free teaching
staff to focus on more creative aspects of the course, including more nuanced
discussions.
Goel and his team began by giving the AI assistant, named Jill Watson to
disguise its identity as a bot, a consolidated memory of all question and answer
pairs from previous semesters. Then they organized the information into
categories based on the types of questions asked. During testing, the team
would ask a question and Jill would have to search previous pairings of questions
and answers for a suitable response.
In January 2016 the professor introduced Jill Watson as one of his new
teaching assistants, and students did not know that this was an automated bot.
 LEARNING AND DEVELOPMENT 145

Like any new teaching assistant, the system offered responses that were
sometimes right and sometimes wrong or strange. For example, a student asked
a technical question about computer languages to use in a project, and the
response from Jill focused on assignment length for a research paper. Goel said it
was much like raising a child, requiring constant support and direction. To help
avoid erroneous support and feedback, the team created a mirror forum,
allowing Jill to answer questions in a simulated environment hidden from
students. If the response was good, a human teaching assistant approved the
message. If not, it was not posted. Over time, performance improved
incrementally, and when accuracy reached 97 per cent for Jill’s responses the bot
was returned to the live discussion board environment among the 350 students.
One of the most appealing value points for the system was responsiveness. A
student in Shanghai could get a response as fast as a student in London or in the
Middle East. Jill enabled responsiveness 24 hours a day, seven days a week, and
365 days a year. Goel’s team actually had to code a time delay into Jill’s
responses so students would not suspect the system was artificially intelligent.
Later in the semester, a curious student finally wondered via the forum if Jill was
an AI or not, and Goel came clean about the experiment. Today, all teaching
assistants for his courses are using pseudonyms and Jill’s name is masked so
new students don’t know which is the bot and which is a real human. The
long-term questions Goel is curious about answering are whether AI really does
reduce teaching staff load, whether it encourages knowledge retention, or if it
impacts student engagement to a measurable degree, but it’s too early to tell at
this stage.

Learning content and curation

Let’s shift gears a bit to content. I’ve said for quite some time that
curation is the new creation for L&D professionals, meaning that
learning leaders need to get comfortable with helping to curate and
find great resources instead of trying to justify their value purely
through the creation of new learning objects. With workers having
the option to search online for virtually any resource, they have less
pressure to visit the employer’s learning management system to find
learning content and assets. That change in pressure is a good thing,
but it also shifts what the employer needs from its L&D team. Instead
146 ARTIFICIAL INTELLIGENCE FOR HR

of feeling as though they need to create learning content for every

eventuality and every unique need, learning teams can prioritize
development of specialized content that they can’t buy or find off the
shelf elsewhere. Additionally, they can put more effort into curating
the best experiences for learners to ensure that content drives engage-
ment and higher-level learning outcomes. As you might imagine, that
curation component can be augmented with artificial intelligence.
Here are a few examples of how that might work.
Historically, the only way for learning teams to find out what
people need is to wait for a request. And by then there’s already a need
and therefore designing, creating and launching the content means the
need may have changed or disappeared entirely by the time the learn-
ing team can respond. But what if there was an advance mechanism to
see what people need before they ask for it outright? One signal for
this would be search volume. Companies could use an algorithm to
capture and record common searches in the learning management
system, company knowledgebase or intranet. If employees are consist-
ently looking for content around a specific topic and it doesn’t exist,
that can spur a discussion about whether that content needs to be
developed. Alternatively, if content already exists on those commonly
searched topics, the firm can curate the best resources focused on
answering those questions and feature them in a more prominent
location, so employees don’t have to work as hard to find them.
It’s no stretch to believe that artificial intelligence should be able to
do this automated curation without human input. When it comes
down to it, people may not even ask for help with something they
need to do their job. I’ve seen in the past where highly qualified indi-
viduals were hired for a high-profile role, and then the individuals
tentatively asked for help and appeared as though they were not fit for
the job. It’s one of those quirky human traits that is hard to predict
from person to person and company to company, because some
cultures are more supportive of collaboration than others. Additionally,
barriers to this scenario may exist because they don’t want to appear
incompetent or unprepared for the job, or it may simply be that they
don’t know the right question to ask of a peer or manager. However,
all of us have been trained with the availability of search engines to
instinctively search for information we need.
 LEARNING AND DEVELOPMENT 147

Another example of machine learning speeding up the learning

curve is by surfacing content that is relevant to you. Not to your
peers. Not to your boss. But to you. Because the learning system
knows your job, your history of completed training, your perfor-
mance records, your informal consumption of learning content (more
on this in the ‘informal’ section below) and your career aspirations, it
can serve up content automatically that helps you to close critical
skills gaps and move you towards your ultimate goals. Instead of
getting a generic piece of content when you log into the system, it
offers to let you pick up where you left off, take a course based on the
next career role you want to take, or strengthen a competency that
you’ve always struggled with. The idea that your learning system
knows you that well is intriguing, but imagine that scaling across a
company of 100,000 people. Every single one of them is receiving
targeted, personalized learning experiences that have been curated to
meet their specific needs.
In addition, any opportunity to upskill workers and improve their
capabilities is a win, so leveraging a bot could offer another option.
We’ve seen throughout this book that bots offer a variety of oppor-
tunities to support personalized interactions with workers at scale.
One solution in the market today is Sunlight. Sunlight is a learning
platform that gives employees ownership over their professional
development. Employers can actually offer this kind of self-directed
learning option as an employee benefit, because a wide base of
research shows that workers are heavily driven by the desire to grow
and develop. Sunlight accomplishes this by helping employers give
each worker their own learning budget to spend as they see fit.
Sunlight has a marketplace of resources and learning assets to select
from but, as we’ve already discussed, the sheer amount of content
can be overwhelming. That’s where the chatbot comes in. Sunlight’s
chatbot offers a concierge experience so workers can converse with
the system to hone in on resources, content and assets that can help
them to achieve their learning and career goals. Most employers
couldn’t dedicate the budget to hire an individual simply to serve in
this ‘learning concierge’ role, but a bot can accomplish it by carrying
out discussions with as many workers as needed around the world at
any time.
148 ARTIFICIAL INTELLIGENCE FOR HR

With more and more learning happening in the moment and being
driven by learner exploration, some say that formal learning is going
to take a back seat to informal learning in the workplace. However, I
always say that nobody ever learned to be a doctor by watching
YouTube. There will always be a need for formal learning to help
guide employees, managers and executives to the right learning and
performance outcomes.

Informal learning applications for AI

Humans are self-developing, self-directed and self-interested. This set
of points forms the core of a talk I give around the myths and truths
of informal learning. In recent years I’ve witnessed a resurgence of
interest in the concept of informal learning, and employers seem to
be very interested in how to not only establish environments where
this happens more freely, but also environments where they can track
the impact and results of informal learning. An exploration of each
of those three points explains the biological imperative to learn and
grow as a human being.
For instance, we are self-developing. We are natural explorers. I
make this point pretty clearly later in the book when I talk about the
skills to future-proof yourself in a more automated workplace. The
curiosity component drives us to ask questions and explore ideas,
sometimes beyond what is even considered rational. We observe,
form a hypothesis, experiment, reach a conclusion and then repeat as
necessary. I always use the joke that this doesn’t happen consciously:
when the first caveman saw his neighbour eaten by a sabre-toothed
tiger, he didn’t verbally hypothesize that he shouldn’t go outside at
night to avoid being eaten himself.
Additionally, we are self-directed. Each of us has a different brain
in a very physical sense. When you learn the history of the capital of
India or how to perform a process in your job better, you store that
information in a different place than I would, even if we both
consumed the exact same type of learning content. That’s because
our brains find ways to connect the concepts to other information
 LEARNING AND DEVELOPMENT 149

already stored in our brain so that it is better retained. Maybe you

visited India so your brain finds ways to make those connections
while you learn about it, but mine instead makes connections to a
news story I read about in the past. Because our brains process infor-
mation differently, we can’t expect all learners to learn the same way.
Finally, we are self-interested. Simply put, we don’t pay attention
to boring things. Think for a second: exactly how many seconds did
it take you to put on your shoes this morning? If you’re like me, you
might not even remember the exact moment when you did. That’s
because your brain eliminates or marginalizes any information it
deems irrelevant. The impact on learning is pretty clear. Do the learn-
ing moments you are creating for your learners rise to the level that
their brains will actually pay attention?
These three components explain the ‘why’ behind informal learn-
ing, but the biggest challenge of informal learning is how to capture
it. As long as people have been alive they have been using compo-
nents of both social and experiential learning. When my three-year-old
son learned to unlock my phone and turn on the drawing applica-
tion, there was no formal course. I didn’t offer an assessment and an
instructor-led lecture on the proper methods and ethics of unlocking
smartphones. He simply watched and repeated the action. In this case
I can observe the outcome of the action, but it’s not always so clear-
cut in the workplace. Smart technologies like artificial intelligence
can help with this issue of tracking and categorizing information that
we learn in the flow of work. For example, what if the learning system
plugged into the collaboration tools employers use? Employers may
have Slack, Microsoft Teams or other systems to capture ongoing
conversations and enable collaboration among employees. How
much informal learning happens among those employees on a daily
basis that could be captured as knowledge? If the bot consumes all of
the questions and answers, it may be able to present answers back
before a human could see the message and respond. For instance, if
James asks Mary how to complete a procedure and she gives him the
response, the next time Alex asks Mary how to complete the same
procedure the system could answer on her behalf, speeding up the
transfer of knowledge and saving Mary’s time for something else
150 ARTIFICIAL INTELLIGENCE FOR HR

more valuable. This is similar to the concept of Georgia Tech incor-

porating a bot into classroom forums to support student questions.
Several years ago I worked with an L&D leader from MasterCard
to share the company’s success story with adopting a new learning
methodology. A core component of the transition was regular video
collaboration among the workers, because it isn’t always cost-effective
to bring learners together into one place for regularly scheduled
sessions. But the challenge with video, even two-way video with
participants, is engagement. When training someone in a classroom,
we can identify whether they are engaged, collaborative, interactive,
and so on. In a virtual environment it can be more challenging to
understand those dynamics. One company that’s trying to sort
through this is onQ. Active learning is more engaging than passive
learning, and that’s one of the core challenges of using video as a
medium – learners are passive. onQ addresses this by scoring engage-
ment, sentiment, facilitation and other outcomes. Imagine being able
to analyse an audience of learners consuming video content to be
able to identify what their general sentiment and engagement were
for the duration of the session? As a trainer and speaker, I would love
to have those kinds of analytics to understand which parts of my
presentations resonated with the audience and which didn’t, in order
to continuously improve the content, delivery and tone.
Learning has always had some prioritization on the content itself,
but in recent years this focus has shifted from an employer-designed
repository of information that employees can visit to a more person-
alized, employee-driven experience. Workers now have the
opportunity to seek out their own learning content and have their
learning technology follow them and record that consumption
outside the walls of the learning management system. For instance,
learning experience platforms like Degreed and EdCast are set up to
capture learning activities on a broader scale. Read an article about
an industry trend? Watch a TED talk to improve your leadership
skills? Those types of activities are tracked and recorded in the
system, offering a more complete picture of your learning.
Every company knows that learning isn’t about completing
courses – it’s about gathering skills and improving performance.
 LEARNING AND DEVELOPMENT 151

Employers are using these learning experience platforms to help

manage employee learning not in the traditional sense but by giving
employees the ability to capture their learning activities on a broader
scale. These tools give L&D leaders a leg up on that challenge. What
truly excites me about these kinds of technologies are their attempts
to focus on outcomes like talent mobility or employee growth versus
just counting the number of courses or articles someone consumes as
a measure of success. Too often, learning sees quantity of content
delivered as its measure of success, when in reality it should be the
improvement of the people and the organization.

Coaching and mentoring

How about coaching? In our study of what separates high-performing
companies from others when it comes to performance management
practices, the top performers are more likely to use performance
coaching as a key component of their talent practices. This is often
used for line-level workers, but managers are sometimes forgotten
about in the shuffle. Tools in the marketplace today that help solve
this problem are dedicated solutions such as LEADx and components
of larger systems like Leadership Actions in UKG’s Perception system.
The core component of both systems is a qualitative analysis of
employee qualitative comments such as surveys and prior perfor-
mance review data. By crunching those inputs the system can generate
recommendations personalized to the managers about how to
improve their leadership style, performance and results.
This concept of a ‘robot coach’ includes the positive aspects of
coaching: an informal, supportive and continuous conversation
about how to improve results. Additionally, because the system can
see actual performance record data, it may be able to give even more
accurate coaching than a human intermediary. And, similar to the
adaptive learning concept explored in the School of One example at
the start of this chapter, these systems should theoretically be able to
see ongoing performance and employee feedback data to understand
whether interventions have been successful at changing behaviour or
152 ARTIFICIAL INTELLIGENCE FOR HR

not. Finally, any trends in the skill gaps of leaders could be tracked
and funnelled to learning teams in order to develop or curate more
targeted resources. For example, if workers are consistently complain-
ing about leadership shortfalls in communications, tools and strategies
can be shared with the management team in a more holistic manner
to address the widespread issue.
Think about this just as businesses think about customer service. If
someone calls a customer support number and is able to be helped by
an automated system, that’s called ‘tier zero’ support, since it doesn’t
require human intervention. If we as HR practitioners can help our
business leaders and managers without having to directly intervene
ourselves, that’s also tier zero support. In other words, it solves the
problem without having to draw us away from the things we need to
be focused on.

Performance support
One of the last components we’ll look at in the learning realm is
performance support. A few things set performance support apart
from other training types: it is accessible at the point/moment of
need, embedded into the flow of work and helps solve a specific prob-
lem. I love the concept of performance support because it supports a
very pure definition of learning: it’s not about the content or the
company, it’s about improving performance. One company that’s
really bringing AI components into performance support is Axonify.
The firm’s system helps employers to codify and replicate best prac-
tices on the job. One retail client saw more than 16,000 per cent
return on investment after it started using the technology. The core
idea is that contextual information in the flow of work is going to be
a game-changer for performance, quality and so much more. Think
about one of the best performers you’ve ever worked with, whether
in the office or outside of it. What did that person do differently from
others that set them apart in terms of performance? Was it a specific
habit? A communication method? A situational awareness? Whatever
it might be, performance support systems will have the capability to
 LEARNING AND DEVELOPMENT 153

capture those differentiators for the best performers and help to

replicate them for the rest of us with contextual alerts, timely recom-
mendations and guided help.

UNDERSTANDING LEARNING CONTENT RECOMMENDATION

SIGNALS

One of the first barriers broken by the vendors serving learning technology
was making content recommendations. The algorithm that powers content
recommendations has become a staple of learning technologies, and it is
powered by a set of common signals. Those signals can be explicit, such as
a user-specified preference or current job title. Alternatively, they may be
inferred, which means the system makes an educated guess about that
user. For example, if I am a graphic designer and I have Adobe Photoshop
on my list of skills, the system may also infer that I am proficient with
Adobe Acrobat or other similar types of applications.
Those signals power the algorithm’s recommendations for content just
the same as your favourite video streaming service does. It knows you and
your habits, and it makes recommendations based on those habits. In the
learner context, the following five signals are some of the more common
ways that learning systems use to recommend content to users.

1 User preferences: By far the easiest to set up in someone’s profile, user

preferences can be tied to job or role. For example, managers take
managerial training. This can also include other defaults or user-defined
preferences such as content about specific aspirational topics or skill
areas of focus.
2 Consumption by similar people: By looking at what similar people take
within the learning management system or learning experience
platform, systems can recommend training that you might also be
interested in that share your current role or interests. If other design
specialists are taking the user experience course, the system can
recommend that you also take the user experience course if you’re also
a design specialist.
3 Consumption by people in my preferred role/career path: This takes the
point above a step further. If you’re currently a marketing associate but
want to be a marketing manager, the system can highlight courses being
154 ARTIFICIAL INTELLIGENCE FOR HR

taken by those at a higher level to help you see what kinds of learning
content you need to succeed in that preferred or aspirational role or
career path.
4 Views (popularity): Recommending popular content isn’t a bad move,
especially if you’re trying to get a critical mass of traffic onto the
learning platform, but this should not be the only indicator of what
content to consume next. We’ve all been sucked into the funny work
video that goes around at work – they are popular but they don’t
necessarily impact performance.
5 Star ratings and comments (value): Unlike raw view counts for popularity,
star ratings and comments can offer a deeper layer of insight into the
quality and value of content. Star ratings, thumbs up/down voting or
other simple measures create a feedback loop about whether content is
valuable or not, and comments can offer deeper insights into the specific
value or feedback points from users.

Looking forward: Virtual, augmented

and mixed reality technologies
To this point, I have avoided talking about the increasing adoption of
virtual reality, augmented reality and mixed reality technologies. It’s not
because I believe them to be unimportant; I just believe their most valu-
able contributions will most likely come in the form of training. One of
my earliest career experiences as an HR professional involved working
for a firm that creates simulation software for military radios, helicop-
ter pilots and students of aviation in college. From those early days I
was able to see not only how the technologies were developed but also
how they actually improved performance for learners. We know that
practice and application are two of the best ways to cement learning
and help workers demonstrate their newly learned skills and knowl-
edge, and for some professions this is more easily done with simulated
environments that virtual and augmented reality offer. Several years ago
I published a case study of HP’s augmented reality printer support tool.
Users could focus a smartphone or tablet camera on a specific printer
and the application would render a three-­ dimensional animation
 LEARNING AND DEVELOPMENT 155

­ verlay on the printer, demonstrating how to change ink cartridges in

o
real-time. This is a simplistic example, but imagine physicians doing this
in a surgical procedure, maintenance technicians on a routine service
process, and so on.
On the virtual reality front, several firms have entered the market
with virtual reality training tools that are claiming very high learning
retention compared to more traditional, passive training methods.
Because of the rise of the #MeToo movement, one area of focus for
virtual reality content has been sexual harassment training. This is
the confluence of several components. For instance, the finding by the
Equal Employment Opportunity Commission that traditional sexual
harassment training doesn’t work helped to shock the industry into
more comprehensive thinking around training for behaviour modifi-
cation, not just to passively share information.11 Additionally, as
mentioned earlier in this chapter, there’s an incredible need to train
workers in more progressive ways to capture their attention and
ensure relevance and retention of the information being presented.
Learning, as we’ve seen here, has much to gain from the advent
and incorporation of new AI technologies. From simplifying the life
of the administrator to delivering better learner experiences, the sky
is the limit for what we can expect to see as the learning profession
embraces automation and machine learning-based systems.

KEY POINTS
●●
Because every learner is different, companies often struggle to offer the
right content and medium that makes sense for everyone. AI
technologies can help to identify the best methods on an individual
level, improving learning outcomes.
●●
A significant portion of learning happens not in a classroom but through
informal learning and collaboration. Artificial intelligence can not only
help to match up learners and mentors, it can also help capture those
informal learning moments in ways traditional technologies can’t.
●●
Looking forward, a blend of virtual, augmented and mixed reality
will help to create more lifelike learning scenarios and situations, to
156 ARTIFICIAL INTELLIGENCE FOR HR

drive active learning. Unlike passive consumption of learning content,

trainers can generate visceral reactions that solidify learning
concepts far beyond the walls of the classroom.

Notes
1 A Rio. Accidental CLO, intentional learning, Chief Learning Officer, 19 March
2018. www.clomedia.com/2018/03/19/jesse-jackson-accidental-clo-intentional-
learning/ (archived at https://perma.cc/DXA4-HJE4)
2 P Galagan. Measurement and evaluation, TD Magazine, 2018. www.td.org/
td-magazine/may-2018-td-magazine-measurement-and-evaluation-issue
(archived at https://perma.cc/8JS2-G8HW)
3 M Corkery. At Walmart Academy, training better managers. But with a better
future, The New York Times, 8 August 2017. www.nytimes.com/2017/08/08/
business/walmart-academy-employee-training.html (archived at https://perma.
cc/6B95-5TF8)
4 B Eubanks. Radically rethinking your learning content strategy, Lighthouse
Research & Advisory, 14 November 2017. lhra.io/blog/new-research-radically-
rethinking-learning-content-strategy-free-report/ (archived at https://perma.
cc/39XR-CUSN)
5 CareerBuilder. The skills gap is costing companies nearly $1 million annually,
according to new CareerBuilder survey, 13 April 2017. http://press.
careerbuilder.com/2017-04-13-The-Skills-Gap-is-Costing-Companies-Nearly-
1-Million-Annually-According-to-New-CareerBuilder-Survey (archived at
https://perma.cc/J2DT-GYSY)
6 M Bidwell. Paying More to Get Less: The effects of external hiring versus
internal mobility, University of Pennsylvania, 2012. https://faculty.wharton.
upenn.edu/wp-content/uploads/2012/03/Paying_More_ASQ_edits_FINAL.pdf
(archived at https://perma.cc/AG2G-N6M8)
7 S Harrington. How international hotel giant IHG is using artificial intelligence
to bring HR and recruitment benefits, The People Space, n.d. www.
thepeoplespace.com/practice/articles/how-international-hotel-giant-ihg-using-
artificial-intelligence-bring-hr-and (archived at https://perma.cc/8W9K-7CMB)
8 B Eubanks. Radically rethinking your learning content strategy, Lighthouse
Research & Advisory, 14 November 2017. http://lhra.io/blog/new-research-
radically-rethinking-learning-content-strategy-free-report/ (archived at https://
perma.cc/3GED-TRX3)
 LEARNING AND DEVELOPMENT 157

9 S Pineau. Learning experience: Collaborative learning and the learning

technologies ecosystem explained – part I, Training Orchestra, n.d. https://
trainingorchestra.com/collaborative-learning-ecosystem-explained/ (archived
at https://perma.cc/85UJ-JKWT)
10 CTPE Communications. Meet Jill Watson: Georgia Tech’s first AI teaching
assistant, Georgia Tech Professional Education, 10 November 2016. https://
pe.gatech.edu/blog/meet-jill-watson-georgia-techs-first-ai-teaching-assistant
(archived at https://perma.cc/HE69-ZBQC)
11 C Folz. No evidence that training prevents harassment, finds EEOC Task
Force, SHRM, 19 June 2016. www.shrm.org/hr-today/news/hr-news/pages/
eeoc-harassment-task-force.aspx (archived at https://perma.cc/3ZGW-KLR5)
158

THIS PAGE IS INTENTIONALLY LEFT BLANK

 159

07

Talent management

In my teen years I was given the task of repairing a piece of our work-
shop’s roof that had been damaged by a storm. I was excited about
the opportunity, because I am probably the least handy person you
can find. If you had to pick a team to build a table or assemble a
bookshelf, I’d be the last one chosen (if anyone picked me at all). Yet
here I had the opportunity to fix something and prove my worth. To
be honest, I was incredibly excited. I would finally undo the years of
being ‘unhandy’ and show off my dormant skills.
I gathered the tools, purchased the material to cover the hole and
drove over to the building. Upon arrival, I realized a critical error on
my part. The ladder would not reach the roof. I looked around the
workshop but could not find a place low enough to climb up, so I
opened the roll up door and drove inside. Bingo! I realized that if I
backed up the truck against the wall, I could put the ladder in the
truck bed and climb out onto the roof. I scrambled up the ladder,
materials in hand. Once I reached the top of the building and got
settled, I went to work. In just a few short minutes I had repaired the
problem, and while it wasn’t as good as new it certainly didn’t have
a gaping hole anymore. I took a few minutes to soak in the glory,
enjoying the beautiful view of the countryside beside me. And then I
realized my mistake.
Remember, I had climbed onto the roof through that hole. The one
that I had just patched closed. I was left with no apparent way to get
down. I debated every option I could think through. Should I hang off
the side and drop, potentially breaking my ankle or leg? Could I call
160 ARTIFICIAL INTELLIGENCE FOR HR

someone to bring me a longer ladder to get down? Maybe I could

remove the repairs I had just made to climb back down the ladder and
return later? After some thought, I took one last look around the edges
of the roof’s perimeter. As luck would have it, one corner had a small
section of roof that extended out about halfway between the roofline
and the ground, which allowed me to ultimately climb down safely,
but this story parallels the conversation we’re having around technol-
ogy in an interesting way. In this example, I created an entirely new
problem (being trapped) by solving an old one (fixing the roof). The
world of AI is not dissimilar. When it solves one problem another may
arise. For example, creating an algorithm to manage worker schedul-
ing, payments and feedback seems like a great way to level the playing
field, reduce bias and improve worker engagement. After all, we’ve
always heard that ‘people leave managers, not companies’, so pulling
the manager out of the process seems like a step in the right direction.
But is that type of work environment really as great as it sounds?
Consider this example of Uber’s algorithm for managing workers.

Replacing humans with algorithms: Hard lessons learned

Uber, one of the most well-known companies in the world, has
become the de facto standard for explaining the gig economy, how it
operates, and its broader impact on the marketplace. The simplicity
of being able to push a button on your phone and have a vehicle
show up to drive you anywhere you want to go within minutes is
truly impressive. The company has a brilliant slogan for those inter-
ested in joining as drivers: ‘Be your own boss.’ In every advertisement
seeking drivers for the company, the messages are positive and seem
to indicate that drivers have control over their environment, assign-
ments and schedule. After all, while it has been a heated debate thus
far, Uber’s assertion that drivers are independent contractors, not
employees, has continued to define the nature of the relationship. For
reference, a few key components for independent contractors that
differentiate them from employees are the options to choose what
clients they want to serve, what hours they want to work and what
 TALENT MANAGEMENT 161

rate of pay they want to accept. With more people seeking flexibility
and additional income opportunities, Uber seems to offer a solution
that allows workers to turn the app on when they want to work and
off when they don’t – the ultimate in flexible work opportunities.
Yet it’s not perfect. Drivers are managed by a faceless algorithm
that tracks their every move, and they often feel coerced into taking
drives even if they don’t want to. But it goes deeper than that. Drivers
have found that they receive different rates of commissions and have
different bonus targets assigned with no insight into how the numbers
are calculated. In addition, some drivers feel like they are being denied
opportunities to drive when they are close to reaching bonus thresh-
olds.1 This lack of transparency seems to be at odds with the claim
that drivers are independent contractors.
However, this isn’t the forum to debate that issue. Instead let’s look
at this idea of being managed by an algorithm and how that affects
the very human drivers at the other end of the employment relation-
ship. A study was performed that examined how drivers are fighting
back against the algorithm in an attempt to regain some sense of their
humanity and control. Drivers use a handful of options, including
switching to other ride-sharing apps or even switching off the app at
critical times, to make their point. According to Paul Hebert, an
expert on motivation and influence in the workplace, this is called
‘negative reciprocity’.2 Negative reciprocity occurs when individuals
act out against unfavourable treatment in an effort to balance the
scales within a system. Hebert explained that this often happens in
less extreme ways in other workplaces, such as when employers have
overly strict dress codes. Employees are always looking for a way to
keep the power from shifting too much towards the employer and
will take actions both large and small to keep things in check.
Take UberPOOL, for example. Unlike the original UberX offering,
which allows a person to take a private ride to their destination,
UberPOOL rides are cheap because they are supposed to allow drivers
to pick up multiple passengers on the same route. While it sounds
good in theory, UberPOOL pay-outs for drivers are lower than the
already inexpensive rates for normal rides, and it adds layers of logis-
tical complexity that can challenge drivers. For example, if my driver
162 ARTIFICIAL INTELLIGENCE FOR HR

stops to pick up two more people after picking me up, I might rate the
driver poorly for a slow trip even though she doesn’t have the option
of not picking up the other riders. Consistent low ratings can hamper
the driver’s ability to get future ride assignments. For these and other
reasons, drivers typically try to avoid answering UberPOOL requests,
but they can be punished for failing to accept rides that Uber assigns
for them. One workaround that some drivers have found is to accept
the first rider on UberPOOL trips and then ignore other requests until
the trip is complete, allowing them to pocket the normal UberX
commission instead of the lower UberPOOL commission.
Additionally, drivers have taken to online forums and social media
sites to organize themselves and find more effective ways to protect
their interests. For instance, drivers will organize mass switch-offs of
the driving application to trigger surge pricing. In areas where there
are more riders than drivers, surge pricing kicks in to balance the
supply and demand of resources. When drivers switch off and remove
the supply, rider demand can trigger surge pricing. Once triggered,
drivers can switch their apps back on by communicating with each
other on back channels and take advantage of the new, higher prices
and corresponding higher pay-outs for drivers.
At the same time, if any issues arise, such as emergency situations
with passengers or simple technical issues with the app, drivers have
a very challenging time communicating with an actual person at Uber
for support. NPR did a study with nearly a thousand Uber drivers to
understand their issues and challenges, finding that nearly eight out
of ten drivers were unable to speak with a human when they needed
help from the company.3 Imagine having an issue at work, whether
emergency-related or simply process-related in your daily routine.
Now imagine not being able to ask anyone about how to solve the
issue. That’s a hint of the reality these individuals face when they sign
up as drivers.
This story is incredibly relevant to this concept of talent manage-
ment because as time progresses I expect to see more companies hoping
to insert algorithms and bots into the management relationship for
employees. From helping to give recognition for a job well done to
reviewing performance scores to providing insight on future career and
 TALENT MANAGEMENT 163

succession opportunities, there is a wide variety of use cases for adopt-

ing artificial intelligence technology in the talent management arena.
While there are ways to do that strategically, this provides a cautionary
tale for what it looks like when employers go too far to the side of
automation. At the end of the day, people want to be managed by
people. There are ‘self-serve’ options for reviewing your own goals,
looking at potential career paths and so on, but we have a fundamental
need to connect with others. Hebert’s reaction to this story was that we
need better human managers, not better HR technology. I think the
right answer is a blend of the two. We need better ways to train our
managers, and some of those were covered in the previous chapter, but
we also need better technology to help managers understand how to
support their people in a variety of situations, from growing their
career to enabling their performance and beyond.

The critical role of managers in talent management

As the example from Uber shows, managing workers is a fundamen-
tally human activity. There are opportunities for AI to overtake some
of the more mundane or laborious aspects of the job, such as setting
worker schedules, but the activities like giving feedback, encouraging
performance and discussing career opportunities and growth aren’t
something we can just hand over to an algorithm. The typical defini-
tion of talent management from a process and technology standpoint
encompasses areas like performance management, succession and
career pathing. However, we’re also going to bring in engagement
and recognition systems since they are inherently tied to that incred-
ibly valuable manager–employee relationship. How valuable?
Managers account for up to 70 per cent of variance in employee
engagement scores.4 In nearly any area of the business, if there was a
single point of failure that was responsible for a 70 per cent swing in
results, employers would be finding ways to add backup systems or
spread the risk, yet for some reason managers are not treated the
same way. That’s one reason I’m particularly excited to see AI tech-
nologies come to this area of HR. Managing others is a tough job. I
164 ARTIFICIAL INTELLIGENCE FOR HR

get it. And yet businesses still operate arcane manager selection prac-
tices such as ‘This person is a great performer, so let’s make them a
manager over other employees’. This type of approach often has
disastrous results, yet it’s still incredibly common.
In the book All In, authors Gostick and Elton talk about an exper-
iment at a large organization where teams with good performance
were categorized as green and those with poor performance were
categorized as red.5 The organization then swapped managers from
teams so that red teams got a green manager and green teams got a
red manager. Within a few months the green managers had elevated
the performance of the red teams to green while the red managers
had diminished the performance of the green teams to red. What this
reinforces is just how significant the influence of managers is on team
performance. We know this anecdotally when we work for good or
not-so-good managers, but there is a variety of data to support the
idea that managers matter to a great extent in the employment rela-
tionship. As in the Uber example, there is a place on the spectrum of
automation that goes too far at taking the humanity out of the
process, but the average company today has a lot of space to run if it
wants to introduce automation and stretch the capabilities of its
managers.
One story I’ve been very interested in focuses on who gives recog-
nition within organizations. A 2016 study analysing employers found
that the givers of recognition matter just as much as the receivers of
recognition.6 One of the interesting findings in this study was that
non-managers give fewer recognitions overall, but those that do
participate actually give them more frequently than managers them-
selves. When I’m in a group of HR leaders and we are talking about
working with our management staff, one of the inevitable issues is
that managers aren’t great at giving recognition or feedback to
employees. What this study shows is that some non-managers may
already have that competency built in, which would potentially be a
signal for moving them into a managerial role at some point. After
all, if these individual contributors are already predisposed to offer-
ing recognition to their peers, why not harness that in a managerial
role? What’s exciting to consider is how we could leverage AI tools
 TALENT MANAGEMENT 165

for analysing how recognition flows in and among the workforce.

For example, if the system sees an individual who is consistently
recognizing peers and others for their work efforts, it might also scan
the person’s performance records to see if they should be flagged as a
management candidate for succession purposes. While it seems like a
relatively simple process, it surely beats the common management
selection practices of choosing the person with the most tenure, pick-
ing the individual who wants the job (even if it’s not a good fit), or
moving the best performer into the position and taking away the job
in which they were performing well.

Employee engagement
Employee engagement is about more than moving some score up or
down by a few points with a survey. I’m convinced that employers
have become ‘engagement weary’ in the last few years, because despite
spending in a variety of areas, little has changed with engagement.
One of the main issues I speak about when I’m discussing this with
business executives is the need to go beyond engagement to look at
actual business metrics and outcomes. There is a variety of research
studies correlating employee engagement levels with things like inno-
vation, safety, quality, revenue, shareholder value, customer
satisfaction, retention and so on. Those are the metrics we need to be
talking about when we discuss engagement, especially if we want
leaders to buy in and support this idea.
An opportunity for AI to support this would be to pull those dispa-
rate measures into a dashboard and show as close to real-time as
possible which were correlated in a particular business and to what
extent. For example, a healthcare firm might see correlations between
engagement and quality more strongly than a financial services firm.
Or, if the connections are there, they may be stronger in one firm or
another because of size, industry or culture. AI would be ideally
suited to examining these connections and correlations to define the
relationship at a firm and how they adjust over time based on changes
in engagement levels. What excites me about this is that engagement
166 ARTIFICIAL INTELLIGENCE FOR HR

could be seen as a leading indicator for key business metrics. While

HR leaders are typically not experts in statistical correlation and
analysis, an algorithm could easily make these connections and then
use predictions to show ‘what if’ analyses based on changes in engage-
ment. For example, if engagement increases by three points, the
employer might be able to expect a 5 per cent rise in sales or a 2 per
cent bump in customer satisfaction scores.
To take it a step further, the system could also work on the other
end of the spectrum to help drive up engagement in meaningful ways.
Because we know managers play a significant role in engagement
variance, the system could suggest ways managers could improve
communications and engagement levels. For instance, in a conversa-
tion with one technology firm that offers both performance
management and learning management technologies, the product
team saw retention data was connected with managers that held
regular one-on-one discussions with their team members. The more
likely managers were to have these meetings with their team members,
the more likely those team members were to stay. Additionally, a
discussion with an HR executive at a large aerospace firm shed some
light on the company’s findings around manager and employee rela-
tions. The team found that managers with larger teams were much
more likely to have issues with engagement and worker satisfaction,
which is a logical conclusion. A manager with three employees should
be able to connect more deeply with the team than a manager with
ten employees. If we combine those two data points, it demonstrates
an opportunity for AI to help guide managers towards more mean-
ingful relationships. Some of the simpler AI applications like smart
notifications could help managers to ensure they stay on top of regu-
lar discussions with their team members. However, more in-depth
tools might give managers insights into which team members might
present a flight risk based on a wide variety of signals and behav-
iours. These kinds of insights are most helpful when accompanied by
prescriptive insights and ideas that can help managers know how to
take the next steps to improve the relationship. Engagement, from a
practical standpoint, is about helping workers to feel like they are
appreciated enough that they give their full effort at work. Algorithms
 TALENT MANAGEMENT 167

can’t make that happen without human intervention, but by enabling

managers to have better insights, information and notifications the
systems can guide them towards better relationships with their team
members.
One of the key issues with engagement is that it’s often measured
by a survey. While there’s nothing inherently wrong with surveys,
they don’t always offer a full picture of the issue that employees are
dealing with on a daily basis. AI tools enable companies to take a
deeper dive into survey data with sentiment analysis technology. I
touched on the broader implications of sentiment analysis in Chapter
3, but one of the clear use cases is around employee engagement. An
example of a firm making use of deeper types of employee data is
Community Bridges. The non-profit organization employs 1,500
people and is based in Arizona. Maddie Nichols, Chief Human
Resources Officer, said the firm has been doing standard surveys for
quite some time to understand employee needs and issues at work.7
As a provider of care services for individuals battling substance abuse,
it’s easy to imagine the wide range of emotions an employee at
Community Bridges might go through in the average day. Yet it’s
important to note that surveys often don’t capture that emotional
component. The firm uses Ultimate Software’s Perception tools to
analyse employee sentiment and emotional state, offering a deeper
perspective into qualitative responses around stress and other recur-
ring issues. Nichols said it’s the difference between getting a vague
‘not satisfied’ response on a survey and understanding the actual,
underlying issue that needs resolution.

CASE STUDY
Understanding employee sentiment in near-real time

What if your organization could understand the top issues for employees across the
organization with just a few minutes of analysis?
Surveys are often carried out quantitatively, because it’s easy to analyse the
responses and come away with trends. For instance, ‘70 per cent of the
organization are engaged.’ Or ‘25 per cent of the firm’s employees are happy
with the CEO.’ The issue is that quantitative data lacks substance and depth,
168 ARTIFICIAL INTELLIGENCE FOR HR

which is where qualitative data comes in. However, the problem with qualitative,
or open text, responses in employee surveys is that they are incredibly difficult
to evaluate and codify quickly. At the same time, these qualitative data points
help provide an additional layer of important information that employers can
use to peel back the hierarchical layers and peer into the organization with
surprising clarity. A great example of this comes from a healthcare client of an
industry-leading HCM technology provider. The firm was facing a series of
common challenges, some of them unique to healthcare and others more
generally focused on the business.
The 10,000-person firm hired a new CEO and wanted him to hit the ground
running. Task number one on his agenda was addressing the entire employee
population in a virtual town hall meeting just a few days after starting on the
job. The issue was the CEO wanted to speak directly to the challenges and hot
buttons for the employees, showing that he was in tune with their needs and
that he was dedicated to meeting those needs to the best of his ability. Instead
of making assumptions about what was top of mind for the employees, the firm
used its technology solution to create a simple survey for employees to
complete.
Within a few days of opening the survey, more than 10,000 pieces of
feedback flowed in through the system. Historically, at this point the data would
either be turned over to a third-party consultant or analysed in-house for trends
and other insights. However, instead of having to manually analyse the data for
trends, hot spots and other friction points, the system (named ‘Xander’ in
honour of Alexander Graham Bell) automatically surfaced the most important
items by leveraging its sentiment analysis and natural language processing
capabilities. All survey responses were factored into the equation, highlighting
key engagement drivers that mattered most to the employee population in near
real-time. The very next day, the leadership team was able to speak openly to
specific employee concerns and avoid further issues with employee morale by
addressing the friction points uncovered by the analysis.
The lesson from this story is pretty powerful. We’re now able to gather and
act upon in-depth data more rapidly than ever before.

I pointed out earlier in the book that one of the unforeseen benefits
of AI tools was helping with the adoption of systems and technolo-
gies. By having bots as an interface mechanism, some companies are
seeing better usage of their tools. However, what if we looked at bots
and AI as a way to help drive research-based best practices in
 TALENT MANAGEMENT 169

e­ngagement, performance and talent management as a whole? By

leveraging the research that exists around the entire employee life­
cycle, AI systems could offer managers, arguably the most important
touchpoint in the employee experience, insights and advice on how
to best drive engagement for their employees.

Performance management and enablement

What makes your performance management process different from
that of your competition in the marketplace? Do those practices you

FIGURE 7.1 Performance practices of high-performing companies

Talent practice Gap analysis: High performers are...

25% less likely to focus on

Focus on eliminating weaknesses
eliminating weaknesses

Forced/stacked ranking 31% less likely to use stacked ranking

Annual goal setting 4% less likely to prioritize annual goals

More frequent goal setting 44% more likely to do more

(two or more sessions annually) frequent goal setting

37% more likely to use recognition

Recognition for performance
to drive performance

29% more likely to use in-the-moment

In-the-moment manager feedback
feedback

26% more likely to use a peer

Peer feedback
feedback mechanism

20% more likely to use coaching for

Coaching for development
development purposes

14% more likely to focus on

Focus on strengths
employee strengths

SOURCE Lighthouse Research & Advisory, 2017 Performance Management, Engagement and
Business Results Survey (n = 259)
170 ARTIFICIAL INTELLIGENCE FOR HR

use actually help enable better performance, or are they processes

that everyone does purely by rote? Research shows that there are
some distinct things that high-performing employers do that other
firms do not, as Figure 7.1 demonstrates.
Each of these practices creates an opportunity to differentiate how
you approach performance management, and the right blend for
your company can improve the employee experience and support
your corporate culture in a powerful way.
On a more individual level, what separates your best performers
from their peers? Can you be specific? In most companies, leaders
may have some sense of what makes a great performer, but it’s not
often spelled out clearly, leaving room for bias. However, even with
clear performance metrics, the playing field isn’t always level for all
parties. For example, a study by Stanford researchers found that
women received less developmental or helpful feedback in perfor-
mance reviews than men.8 In the research, women were more likely
to receive vague praise while men were more likely to receive helpful
feedback on how to improve their performance. Additionally, women
received more criticism around communication styles than any other
job component, and comments for being ‘too aggressive’ in commu-
nication happened to females three times as often as males. What’s
intriguing is that male managers didn’t look very positive in the find-
ings from this study, but when female managers were examined, they
doled out their feedback more evenly to both men and women.
Whether that’s because they had been on the receiving end of that
unhelpful feedback in the past or simply because they are less likely
to be biased in their approach, it’s interesting to see that this phenom-
enon is not universal for all leaders.
These kinds of findings clearly indicate an opportunity to incorpo-
rate artificial intelligence into the equation. What if an AI-based tool
could warn managers when they are giving consistently lower reviews
to women rather than men? Or to older workers? Or to diverse work-
ers? When I worked as the head of human resources for a technology
startup, one of my managers at a remote site was very clear in his
 TALENT MANAGEMENT 171

dislike of one of our employees who was also in the military part-
time. Because of the legal protections provided by the law, he could
not punish this person when they required time away from work to
perform military duties. At the same time, he looked for every chance
he could get to try to ‘get even’ for what he saw as a thorn in his side.
I had to manually examine things like performance reviews to make
sure he wasn’t making overt or veiled comments about this employ-
ee’s time away from work because we were not legally able to utilize
that information in our reviews of the person’s performance. While it
might not be something you’ve dealt with personally, this isn’t an
isolated incident and it didn’t only happen at my firm.
Consider a secondary process that connects to performance for
many firms: compensation reviews. Every year in my role as the HR
leader I scheduled a compensation review board meeting between
myself and the founders of the company. Once performance review
and salary increase data came back from across the organization, we
blocked off time to look at the data from a compliance and bias
standpoint. Were men consistently given raises in the top 50 per cent,
or was it evenly dispersed? What about other diverse employee
classes? This exercise was, to be frank, tedious. Yet it had to be done.
When there were discrepancies or issues, such as when the top 40 per
cent of raises were given to men while women were scattered through
the bottom 60 per cent, we had to take steps to correct the issues and
balance the scales appropriately. While I’m proud of the fact that we
had no complaints or challenges on pay discrimination in my tenure
there, I would have been just as happy turning this process over to an
algorithm. An AI-based system might have caught things I didn’t,
such as a manager consistently underrating females on a yearly basis,
a subtle departmental issue that affected a minority group of work-
ers. With proper calibration, a system like that might have even been
able to recommend adjustments to ensure pay parity among the
employees instead of relying on a group of three individuals to manu-
ally modify the manager’s suggestions.
172 ARTIFICIAL INTELLIGENCE FOR HR

COMPENSATION DATA CREATES RICH ARENA FOR AI USAGE

In an overview of the compensation technology and data provider

landscape early in 2021, our team looked at nearly a dozen different
providers of technology and data to support compensation activities within
the business. We found an increasing number of applications and use cases
for artificial intelligence within the compensation functions of HR. The
providers below are leveraging the immense data sets in compensation,
often made up of millions of individual data points, to create smarter
processes, anticipate needs and generate new value:
●●
Salary.com: Find employee retention risks and get notifications and
alerts for human review and intervention, as needed.
●●
PayScale: Get recommendations for data or pricing patterns based on
the activities of similar companies.
●●
Willis Towers Watson: Use real-time data to determine pricing impact of
skills on a particular job.

Just a few years ago, none of these types of capabilities was even possible
in the compensation technology providers we looked at. This is just a brief
look at how this part of the technology landscape is evolving alongside
other aspects of talent and HR.

Much of what has been covered around performance so far has been
about preventing biases, yet there’s so much more that can be done to
actually enable better performance for workers. One of our firm’s
research studies found that 74 per cent of high-performing firms have
a performance management practice that improves engagement levels
while low performers are 58 per cent more likely to have an ineffec-
tive approach that may actually hamper employee performance.9
How astounding is it that some firms have created performance
management processes that prevent employees from being able to
perform at their best? It’s funny-yet-not-funny to think about the
process we as HR leaders have developed for performance manage-
ment. When I speak about performance management the audience
almost always laughs at the general idea that employees and managers
 TALENT MANAGEMENT 173

universally hate the process, yet the truth is not far from that state-
ment. In most firms the process of performance management has
nothing to do with the intended result. Ask any HR leader and they
will tell you that performance management at their firm is designed to
help track worker performance and help them improve their results,
yet virtually every system I’ve run across fails to do the second half of
that. For sure, we’re great at tracking performance in the past. We can
talk about someone’s past performance all day long. But when it
comes to actually enabling them and helping them to be better
performers, we stumble. AI technologies can help us to improve this
particular aspect of the process, supporting managers at key moments.
For example, one of the challenges mentioned previously is the
manager’s daily struggle with offering in-the-moment feedback to
employees. AI tools can help to keep this practice on a manager’s
radar so they don’t miss opportunities to offer feedback and coach-
ing to employees. While not marketed as a performance management
system, the Achievers Listen product includes Allie, a bot that reminds
managers that they need to offer feedback to their teams. In addition
to reminders, Allie can ask questions, gather and analyse feedback
data, suggest action steps, and more. Because it’s built right into the
flow of work by being integrated with systems like Slack and
Microsoft Teams, managers don’t have to go to a separate system to
offer feedback to employees. When I think about one of the biggest
challenges of getting managers to adopt a system or process for
performance management, the hurdle in many cases comes down to
getting them to use another tool that isn’t in their daily workflow.
From a performance management and feedback perspective, the
more we can integrate this process into the ways and channels the
manager already uses on a daily basis, the more likely it will be
adopted and utilized. Additionally, as we’ve already said within the
book in various places, chatbot notifications can be great nudges for
managers to do the things they already know they need to do. I’ve yet
to meet a manager who says, ‘I really don’t want to give feedback to
my employees.’ Yet they often don’t. If we can help them and keep
this top of mind, they’re more likely to participate, and everyone
benefits from those kinds of behaviours.
174 ARTIFICIAL INTELLIGENCE FOR HR

In addition, I’ve seen a few tools in the market that solve a very
common challenge around performance feedback. You may have seen
this scenario before yourself: a manager gives an employee a rating of
two out of five, but feels bad about it. To compensate for those feelings,
the manager gives a nice, glowing review of the person’s personality,
making them sound like a wonderful employee. Then, two months
later, the manager wants to fire the person. I’ve seen it in nearly every
company I’ve worked for, and it usually gets a knowing laugh from the
audience when I’m onstage because we all know it’s true. One of the
newer capabilities in some systems, like the one offered by
WorkCompass, actually rates the sentiment or feeling of the words the
manager uses in feedback. If the score is a two out of five and the feed-
back is 100 per cent positive, then the system will flag the review and
prevent it from being submitted until the manager corrects the issue
and brings the rating or the commentary in alignment with the other.
Taking it a step further, WorkCompass also helps managers to craft
better and more helpful feedback overall, enabling the performance
process to actually improve performance as it is designed to do.
One of the final components of performance discussions that ties
in nicely with the next section is around career development. Workers
want to grow and develop, and if they can’t envision a future path at
your company they will surely envision it somewhere else. Managers
need to have conversations with workers about what kinds of devel-
opment they need, what tasks they enjoy and what career aspirations
they might have, but those kinds of things are often lost in the shuffle
of tasks and priorities that change from day to day. Managers need
to be open to this kind of conversation, but employees really need to
drive it. AI technologies can help to bring these discussions to the
forefront in a variety of ways, as we’ll explore below.

The gig economy and team development

In 2013, World Bank Group was in trouble. The firm’s leader, Jim
Kim, had just put plans in place publicly to cut more than $400
million in the following decade, leading to a variety of restructuring.
 TALENT MANAGEMENT 175

One of the key components of this cost cutting? A hiring freeze. The
challenge of a hiring freeze is that while new workers can’t be brought
in, work still has to get done. An innovative solution to this problem
was the launch of SkillFinder in 2014.10 Prior to the launch, there
was no way for leaders to drill down into the skills of the 27,000
workers across the firm, forcing them to rely on things like job titles
to guess what skills an employee might possess. However, the limita-
tions of that approach are fairly evident: a project manager for a
software development team may have wildly different skill sets to a
project manager for a nutrition-focused project in one of the compa-
ny’s non-profit endeavours.
The launch of SkillFinder was focused on helping to create better
teams and visibility into employee capabilities across the enterprise. I
liken it to an ‘eBay for talent’, where workers can offer up skills and
individuals from across the organization can put up projects that
require diverse skills to complete. This was the initial goal and
purpose, but leaders quickly realized the system was being used in a
way that they didn’t foresee. In the first year of use, a flurry of activity
focused not just on building teams but on exploring and discovering
the skills of peers. According to the firm, ‘Managers and employees
want to understand better the skills and skills gaps of those who sit
across from them in order to inform their work plans.’ This internal
‘expert network’ had been co-opted by the employees in order to help
grow their understanding of the skills and capabilities of existing
team members.
While this program was launched without the aid of machine
learning, it’s a great example of a type of process that AI could accom-
plish better than a human, because in spite of the successes there are
still humans doing the matching and selection. One firm that has
created technology to solve this problem for a range of employers is
Nexus AI. The technology helps to surface the best employees for a
specific team based on the purpose of the team and the skills of the
employees themselves, and it can even tap into external resources like
freelancers and contingent workers, if needed, to supplement the
capabilities of the internal workforce.11 According to one interview
with the founder of the firm, Nexus AI looks beyond the basic details
176 ARTIFICIAL INTELLIGENCE FOR HR

most project management systems would consider, like someone’s

skills or their available time and schedule. It also examines dynamics
around team interactions and organizational behaviour attributes.
For instance, is the leader of the team creating the right environment
of psychological safety? Consider also the value of simply having the
insights generated by a matching system like this. Employers can
understand the volume of requests for a particular skill in a holistic
way, potentially seeing trends that could be addressed with training
or additional hiring activities.
Systems such as these are unaffected by favouritism and other
human biases, which means the matching can happen more quickly
and potentially improve things like team diversity and performance.
The case study at the end of this chapter demonstrates how Shell uses
a similar system, Catalant, to solve its challenges with rapidly devel-
oping new teams and making use of the deep, yet dormant, skills that
many of its workers possess.

Internal talent mobility and career pathing

When I think about internal mobility, the first company that comes
to mind is a particular financial services firm. In an interview with the
company, I realized that the firm was in trouble; its struggles also
represent the similar issues experienced by a significant portion of the
marketplace today. The firm was in an uphill battle with its talent
management and talent mobility efforts. In the discussion with the
digital services leader for the financial services firm, I learned that of
the nearly 50,000 employees, the company had clarity into the actual
skills and competencies for about only 10 per cent of those people.
That left approximately 40,000 workers who were nothing more
than a job title on an organizational chart because the firm didn’t
have granular information about the skills and capabilities of each of
those individuals. This lack of clarity was a challenge, because this
particular leader had been tasked with developing specialized teams
using the existing workforce. This went beyond not knowing how to
tap into those employees for team development – it also presented
 TALENT MANAGEMENT 177

issues with recruiting and learning. If you don’t know the skills your
people have, how can you target gaps with training? Additionally, if
you have open roles, how can you promote them to the right people
if you’re not aware of the capabilities of your existing workers?
The truth is that employees who find another opportunity within
your company are more likely to stay, yet employers are notorious
for a lack of transparency around job openings internally. One
research study found that more than half of employees said their
company made career opportunities known somewhere between
never and occasionally, and only a minority said it was done on a
regular basis.12 This is an issue, because in our extensive analysis at
Lighthouse Research of internal talent mobility, we’ve uncovered a
wide variety of use cases, examples and stories that help to illustrate
the value of internal mobility.
Chipotle, a fast-casual restaurant chain, reduced store manager
attrition by more than 50 per cent when it switched its practice from
hiring managers externally to only hiring managers from within the
ranks of the existing employee population.13 Credit Suisse has a
programme called Internals First, where recruiters at the company
call their own employees about job opportunities before looking to
external candidates, which has led to the company saving more than
$75 million in hiring and training costs.14 When we add to this data
from the Wharton School about how employees that are hired from
outside the company cost approximately 20 per cent more money but
perform worse on average for their first two years on the job, it
becomes clear that hiring from within the ranks of the company is a
powerful way to drive better performance and results.15
Several technology vendors sell systems that help to solve this
problem. For example, one firm’s system guides workers through the
process as they look for career opportunities within the firm, helping
them get up to speed on what the new job requires. It all starts when
employees put in information about their interests, skills and aspira-
tions. Then the system identifies opportunities that match the unique
skills fingerprint for each employee. As workers look at opportuni-
ties, the system can highlight the gaps between current and future
178 ARTIFICIAL INTELLIGENCE FOR HR

skills in the new career track along with actual learning resources to
fill those gaps. The learning component can be actual formal learning
content or informal and social learning relationships with mentors
and subject matter experts. Think about it from the mentoring and
social learning aspect: being able to see the gap between your current
and your needed skill set is one thing, but being able to have conver-
sations with subject matter experts and top performers in that job is
an exciting prospect. Some technologies even allow workers to work
on smaller projects in an effort to stretch and grow their less-developed
skill sets. All of these interactions are powered by AI and are scalable,
which means that the system can be helping individuals across the
firm create their own unique growth and development plans. While
at the end of the day managers are going to be involved to support
the transition, onboarding and performance components, the finer
details of these interactions can be easily managed by a smart
­algorithm.
Leadership succession is one component of internal mobility that
is clearly affected by the advent of AI technologies. If we translate the
common issue with bias in hiring selection to succession selection, it’s
pretty obvious that employers will not make strides in increasing
diversity in key leadership roles without major interventions.
However, if those decisions are initiated or supported by unbiased
algorithms, employers can find better, more diverse candidates for
succession than the common method of ‘gut feel’ that employers
default to. Additionally, because AI systems can use a wide variety of
inputs, it might be able to see patterns that humans can’t. For exam-
ple, a manager might think a candidate is qualified for a leadership
role when in reality the system can see that the individual’s perfor-
mance suffered in the past when they transitioned to leadership
positions, signifying that the person may be better suited to an indi-
vidual contributor role. Instead of forcing someone into a role that
might not fit, the AI can help to guide decisions using a more full and
accurate picture of their overall performance history.
And just in case you thought we’d get past this topic without
bringing up chatbots, there are employers that use bots to help guide
 TALENT MANAGEMENT 179

employees through these types of conversations about career aspira-

tions. Just like we discussed in Chapter 5 about how bots can help
with candidate questions in the talent acquisition process, bots can
also assist internal workers looking for that next opportunity.
However, the interactions can include even more data and insight
into existing employees and career paths since the system user is
already an employee. IBM uses its own tools in-house to help employ-
ees understand what career opportunities exist, the degree to which
their skills match the openings, and what other individuals currently
hold those types of positions.16 The idea that I could, in a company
of nearly 400,000 workers, find other individuals across the enter-
prise that hold jobs I might be interested in is intriguing. The team at
IBM believes that managers simply aren’t equipped to have transpar-
ent conversations about career opportunities. Even if the manager
wasn’t worried about losing and having to replace a critical employee,
which is a common issue, there is a clear trend in individuals wanting
to take ownership of their careers. Even if this means moving beyond
the typical career path or career ladder to more of a career ‘lattice’
structure, workers want some measure of control.

CASE STUDY
Algorithms determine teams at Royal Dutch Shell

One of the most common struggles for employers is understanding the breadth,
depth and variety of skills across the employee population. Understanding your
own team is one thing as a manager, but trying to guess at the skills of hundreds
or thousands of employees across the enterprise is another issue entirely. Some
companies rely on job titles as a map into what capabilities employees have,
using titles as a proxy for skills. So if someone’s job title is ‘project manager’ then
we could assume the person is proficient in use of project management software,
budgeting, scheduling and so on. But what if the person’s prior job was working
as a marketing executive or an operations leader? Those kinds of details don’t
show up in a job description and without a smarter, more granular method for
mapping worker competencies, those kinds of insights would go unnoticed.
180 ARTIFICIAL INTELLIGENCE FOR HR

In the past, when firms wanted to set up a new team to pursue a project, they
had to search the employee population based purely on first- or second-hand
information to try to identify the right mix of skills and capabilities. The bigger
the organization, the more difficult it would be to find the right people. For Royal
Dutch Shell, a global oil and gas company with more than 90,000 employees,
that would present quite the challenge. So the firm turned instead to an
algorithm to help establish the right teams.17 Shell uses Catalant, a system that
helps employers to develop internal talent marketplaces where workers can
define their skills using in-depth profiles. Conversely, project leaders can lay out
initiatives they need help with so workers have the opportunity to share their
expertise in areas that interest them. Shell’s adoption of the system started in
2017 and it rolled out the Shell Opportunity Hub within the firm’s business-to-
business marketing division with nearly 8,000 employees.
Even as recently as a few years ago, employers would have to reach into the
open market and hire an external candidate if they were unable to identify the
right internal talent to pull into strategic projects. While there’s nothing wrong
with hiring to staff a team, this often meant qualified, available and interested
internal candidates were passed over simply because of a lack of clarity around
capabilities or aspirations. Shell knows it has a deep bench of employees on tap
who are willing and able to carry out the projects it has – if only it can find them.
Caroline Missen, Business Advisor at Shell, said, ‘We’re looking at how we can
efficiently access and use the diverse talent we already have in Shell.’18
Systems like Catalant help employers to manage the process of creating new
teams, but they can go further as well. For example, these types of software can
help with scheduling and guiding strategic projects through milestones, a
process previously reserved for human project managers. One use case for the
system at Shell was when the firm needed help looking at digital business
models in the car maintenance division of the business. With a few clicks the
system scanned available talent within Shell and assigned them to the team.
Another issue with project-based work? A lack of clarity around performance.
If workers are not participating in a team with their manager, it is often difficult
to get data on how the individuals performed as a part of the initiative or
project. The tools Shell uses help to capture that information and ensure a level
of transparency around performance that helps with future team development.
For instance, if Mary lists a specific skill or expertise in her employee profile but
 TALENT MANAGEMENT 181

fails to perform on a project, the algorithm can refine the search results to move
other candidates higher up the list in future matching attempts.
As we’ve already seen, there’s a fine balance to the influence and authority of
people and algorithms in a work situation. The task of project management is
heavily process-oriented and therefore a good opportunity to leverage the
strengths of an algorithm to select teams and assign tasks without bias in
addition to gathering performance data. At the same time, the algorithm is
looking for the strongest and best-fit capabilities of the individual employees to
predict and create the best-performing team possible. This is a great example of
how to blend the best of human and technology to create better outcomes for
the people and the business.

Because talent management touches on so many facets of the employ-

ment relationship, the integration of AI into the mix creates ample
opportunities to serve up personalized information, recommenda-
tions and more. Having an employee’s performance data on hand is
one thing, but being able to also see where they best contribute to
teams, how engaged they are relative to their peers and what succes-
sion opportunities exist creates a more complex and complete picture
for employers. And as I’ve said previously, AI is about helping us to
make better predictions and decisions. One misstep in the talent
management function might be all it takes to lose that star performer
to the competition.
In the last few chapters we have covered some of the most impor-
tant tasks that HR leaders are responsible for: core HR, recruiting,
learning and now talent management. But we’re not quite through.
The next chapter takes a step back from the day-to-day activities and
shines a strategic spotlight on areas where you can expect to see chal-
lenges in the adoption and usage of AI technologies. While I am an
optimist and believe we’re better off with these tools at our disposal,
that doesn’t mean this monumental transition won’t occur without
its own set of challenges. Consider this an opportunity to pre-empt
the inevitable issues by educating yourself thoroughly on critical
areas like algorithm aversion, data privacy and more.
182 ARTIFICIAL INTELLIGENCE FOR HR

KEY POINTS
●●
Employers are understandably excited about the capabilities offered by
algorithms, but we must be careful not to automate too much, as
removing humanity from some processes can lead to backlash in the
employee population.
●●
Managers play a key role in the engagement of employees. Having AI
technology on hand to support that relationship, not replace it, is where
the true value lies.
●●
Business leaders can use AI to create a more holistic view of employees
and their capabilities, leveraging that information to rapidly build teams,
create talent pipelines for key roles and solve business-critical problems
with the right mix of skills.

Notes
1 P Solman. How Uber drivers game the app and force surge pricing, PBS, 4
August 2017. www.pbs.org/newshour/economy/uber-drivers-game-app-force-
surge-pricing (archived at https://perma.cc/8QEY-C87T)
2 P Hebert. Your biggest employee engagement challenge is after the economic
recovery, HR Exchange, 1 December 2010. www.hrexchangenetwork.com/
hr-talent-management/articles/your-biggest-employee-engagement-challenge-is-
aft (archived at https://perma.cc/82H2-9XTG)
3 A Shahani. The faceless boss: A look into the Uber driver workplace, NPR, 9
June 2017. www.npr.org/sections/alltechconsidered/2017/06/09/531642304/
the-faceless-boss-a-look-into-the-uber-driver-workplace (archived at https://
perma.cc/9NHD-Z6RM)
4 R Beck and J Harter. Managers account for 70% of variance in employee
engagement, Gallup, 21 April 2015. http://news.gallup.com/businessjournal/
182792/managers-account-variance-employee-engagement.aspx (archived at
https://perma.cc/YR7T-YEXC)
5 A Gostick and C Elton (2012) All In: How the best managers create a culture
of belief and drive big results, Simon and Schuster, New York
6 G Beckstrand, J Rogers and D Sturt. Influencing Greatness: Giving, receiving,
and observing recognition, OC Tanner Institute, 2016. www.octanner.com/
content/dam/oc-tanner/documents/white-papers/InfluencingGreatness
Whitepaper.pdf (archived at https://perma.cc/P6NT-HRS8)
 TALENT MANAGEMENT 183

7 D Marcroft. Growing non-profit healthcare provider uses UltiPro perception

to build trust, support employee success, Ultimate Software, 16 May 2018.
www.ultimatesoftware.com/PR/Press-Release/Growing-Non-Profit-Healthcare-
Provider-Uses-UltiPro-Perception-to-Build-Trust-Support-Employee-Success
(archived at https://perma.cc/YV4G-D3AF)
8 SJ Correll and C Simard. Vague feedback is holding women back, Harvard
Business Review, 29 April 2016. https://hbr.org/2016/04/research-vague-
feedback-is-holding-women-back (archived at https://perma.cc/L9YV-2DJR)
9 Upstart HR. Moving beyond management, Lighthouse Research & Advisory,
November 2017. https://upstarthr.com/wp-content/uploads/2017/11/LH-High-
Value-Performance-Management-Practices-Graphic.jpg (archived at https://
perma.cc/K8SW-3C3P)
10 GovLab. The World Bank: Skillfinder, 10 February 2016. www.thegovlab.org/
static/files/smarterstate/skillfinder.pdf (archived at https://perma.cc/9EUS-R626)
11 A Chowdhry. How Nexus AI is helping companies discover untapped talent,
Forbes, 13 November 2017. www.forbes.com/sites/amitchowdhry/2017/11/13/
nexus-ai/#65c561b94012 (archived at https://perma.cc/NJ4Y-ATUT)
12 S Bronstein. One in three employees claim to have a job rather than a career,
new Mercer survey finds, Business Wire, 12 August 2015. www.businesswire.
com/news/home/20150812005906/en/One-in-three-employees-claim-to-have-
a-job-rather-than-a-career-new-Mercer-survey-finds (archived at https://perma.
cc/D2RR-RRNG)
13 Upstart HR. Internal promotion: How Chipotle reduced turnover by 64%,
North Central Florida Society for Human Resource Management, 14 October
2011. https://ncfshrm.shrm.org/blog/2011/10/internal-promotion-how-
chipolte-reduced-turnover-64 (archived at https://perma.cc/YH8L-NE2S)
14 P Garland. Why people quit their jobs, Harvard Business Review, September
2016. https://hbr.org/2016/09/why-people-quit-their-jobs (archived at https://
perma.cc/7VZX-4Z85)
15 M Bidwell. Paying more to get less: The effects of external hiring versus internal
mobility, Administrative Science Quarterly, 27 December 2011, 56 (3).
http://journals.sagepub.com/doi/abs/10.1177/0001839211433562 (archived at
https://perma.cc/4UM7-CMMV)
16 https://www.shrm.org/resourcesandtools/hr-topics/technology/pages/ibm-the-
future-of-ai-hr-closing-emerging-skills-gaps.aspx (archived at https://perma.
cc/484J-MLKB)
17 S Schechner. Meet your new boss: An algorithm, The Wall Street Journal, 10
December 2017. www.wsj.com/articles/meet-your-new-boss-an-algorithm-
1512910800 (archived at https://perma.cc/4KQT-Z34U)
18 S Schechner. Meet your new boss: An algorithm, The Wall Street Journal, 10
December 2017. www.wsj.com/articles/meet-your-new-boss-an-algorithm-
1512910800 (archived at https://perma.cc/G2LD-MLGT)
184

THIS PAGE IS INTENTIONALLY LEFT BLANK

 185

08

Challenges of adopting
AI technology

If you go to the Google Trends feature, you can see the history of a
term and how popular it has been for searches. Ten years ago, virtually
nobody searched for or had even heard of the concept of biased artifi-
cial intelligence. Then, in 2019, the searches for the term skyrocketed
when a story broke in the news about one of Amazon’s in-house algo-
rithms its team had developed. While the story has been used in a
variety of contexts since then to call out Amazon for poor form, the
truth is that the company brought the conversation to the forefront,
paving the way for many researchers to dive in and explore the topic
more fully. In many of the speaking engagements I have carried out
since then I’ve had audience members ask about bias in algorithms,
citing the Amazon story as a source. If you haven’t heard it, here’s the
full story along with the lessons learned for every talent and HR leader.
One of the struggles that Amazon has, like every large employer, is
hiring the right talent. If there was a way to speed up or automate
some of that hiring process, it could save millions of dollars and
hours over time, so the company tasked one of its internal teams with
building an algorithm to select the best people to hire. The algorithm
was based on the existing workforce at Amazon, which is predomi-
nantly male (as with many technology firms). When the algorithm
produced its results, it primarily highlighted resumes of men. It not
only presented men as more favourable candidates, but it also down-
graded anything on a woman’s resume that indicated her gender, such
186 ARTIFICIAL INTELLIGENCE FOR HR

as ‘women’s chess club’ or the fact that she had graduated from an
all-female university. Even if the software was told to ignore those
factors, it could still look for other proxy data to indicate gender.
Amazon claims it never used this as a method to pick candidates,
but the lesson remains: the way these tools are built can fundamen-
tally set them up for success or failure. Within this chapter we’ll
explore bias as a concept more fully, detailing other examples to help
you understand how an algorithm built to solve a problem might end
up causing more problems down the line.
This is a challenge, of course, but there are other challenges that
we must face as we attempt to bring more AI into the workplace.
Those challenges include everything from a general mistrust of data
collection and algorithms to more theoretical questions like when
humans should step back into decision-making and how we will
know to retake control when algorithms begin making more impor-
tant decisions in the workplace.

Biased systems
Algorithms run on data. In one experiment profiled by the World
Economic Forum, researchers had an algorithm parse over three
million books published between 1900 and 2008 and then describe
men and women based on that analysis. The algorithm proceeded to
describe women primarily by their looks, including words like ‘beau-
tiful’ or ‘classy’. The men? They were described based on their
behaviours, with descriptors such as ‘brave’ or ‘righteous’.1
Any one of us could immediately see the inherent issue in describ-
ing men and women using different types of words, but the algorithm
couldn’t see its own faux pas. This matters in the broader conversa-
tion around AI in the workplace, because machine learning algorithms
are trained on immense amounts of text, and the sources we provide
can bias those findings from the very beginning.
In the 2017 Artificial Intelligence Index report, one of the points of
analysis was around sentiment of media coverage around AI by year.2
In 2016, negative coverage of AI spiked to the highest point to date.
 CHALLENGES OF ADOPTING AI TECHNOLOGY 187

If we set aside all of the ‘AI and robots are coming for our jobs’
hyperbole, one of the most common concerns with artificial intelli-
gence is perpetuating bias using algorithms. For example, the ‘fake
news’ that crops up on social media is delivered to us based on our
own preferences and habits. If I already believe the sky is purple or
that a particular politician is untrustworthy and I read about or share
those topics on social media, the platform I’m using will offer up
even more examples and sources of information confirming those
types of biases, further shutting me off from a balanced view of the
world. The bias I already have is being reinforced and further deep-
ened by algorithms giving me what I want. This personal example
may seem relatively innocuous, but there’s a darker side that impacts
the workplace as well.
One day a Harvard professor was working on a project and
Googled her own name to find something. However, she was shocked
by what appeared in the sponsored advertisement section of the page.
In the little box she read, ‘Latanya Sweeney Arrest Record.’ Knowing
that she’d never had an arrest record, she clicked through to see what
it led to. Ultimately it was just another pay per click advertisement on
Google’s search results, but it piqued her curiosity.3 In the following
months, Sweeney conducted a research study to understand how and
why the ads appeared, and realized through her experiments that
‘black-sounding’ names were much more likely to return an ‘arrest
record’ result in a search while ‘white-sounding’ names were more
likely to return more benign results. Remember, computers only use
what they are programmed with, and if more people are clicking on
the alleged arrest records then the algorithm will prioritize those in
search results pages. These types of algorithms reinforce what we
already believe, whether it’s correct or not.
Another side of this is making predictions based on biased algo-
rithms. For example, in the chapter on talent acquisition I talked
about using AI as an unbiased tool for matching people to job open-
ings. We should expect this to be fair, right? It doesn’t take into
account gender, race or age when matching a candidate’s resume or
skills with a job advertisement. Yet it’s possible that those predictions
and suggestions the algorithm is making are entirely biased, d ­ epending
188 ARTIFICIAL INTELLIGENCE FOR HR

on the models used to train the system. Consider this example: you
work for an investment bank that draws a significant portion of its
workforce from prestigious MBA programmes at Ivy League univer-
sities. You use an algorithm that is designed to find candidates that
match the high-performing workers you already have. If those work-
ers all have similar backgrounds like Ivy League schools, specific zip
codes or other components, you may be discriminating against
minorities and women in your selection process. Think about it for a
second. The system’s goal is to screen a batch of candidates and look
for individuals similar to those that already have jobs at your firm. If
you have an overwhelming population of young, white males in your
firm, the system will try to select that same population.
It’s a bit unsettling, but it also can serve as a reminder to have
regular check-ups on our existing processes and practices (whether
AI-powered or not). It’s also a reminder that humans can never fully
let go of the reins (more on this in a following section). If our current
approach is biased to some degree, then we need to take steps to
change it, and the first step is to identify whether the approach is
biased or not. Early in 2018 Pymetrics, an assessment company
focusing on the talent acquisition market, open sourced its AI tool for
detecting algorithm bias.4 Companies can use this tool to identify
whether an algorithm is selecting based on biased factors or not.
While we would hope that vendors in the space would be accounting
for these kinds of things, that isn’t always the case. Tools like the one
Pymetrics released will help to solve these issues with biased algo-
rithms, however common they may be.
Bias, ethics and decision-making are tightly linked when it comes
to AI. During an interview with Sarah Andresen, Principal People
Scientist for Sage AI Labs, she highlighted the critical importance of
ethical artificial intelligence decisions. Andresen says there are three
areas employers need to consider before using AI: what data was
used to train the models, how the technology will be used and what
the ethical implications might be. In her experience as an AI expert
and HR researcher, AI is only as good as the data it is built upon and
it should always be used ethically. It is important for companies to
consider the quality and completeness of training data along with
 CHALLENGES OF ADOPTING AI TECHNOLOGY 189

any biases that may exist in the data. In addition, companies should
consider the ethical implications of AI and how it will be used, ensur-
ing they are transparent in its application.
An important point to remember is that AI does not inherently
understand the concept of bias. The algorithm runs an operation, comes
to a conclusion and presents that information. It’s up to us as develop-
ers and implementers of AI solutions to be aware of bias and where it
might rear its ugly head. In one interview with IBM’s Distinguished
Engineer Lisa Seacat DeLuca, she explained that IBM attempts to miti-
gate bias issues during the design process by assembling teams of diverse
individuals to create each algorithm. In theory, the more diverse the
group, the less likelihood that the team will build something that has an
adverse impact on any particular gender, race or other group. Biased
algorithms are a real danger and one that we must be vigilant about as
more AI technologies enter the workplace. After all, the worst that can
happen to the system is to be shut down. However, the human in charge
of making the decisions based on the system’s recommendations can
lose their job if the errors and decisions are grossly biased and unad-
dressed by the ultimate decision maker, resulting in unfair and
discriminatory employment practices.

BALANCING AUTOMATION AND FAIRNESS

In a study released in early 2021, our team identified how 856 different
companies are trying to balance the increasing demand for automation in
hiring with a focus on fairness and equity.5 The most common method?
Assessments and other tools designed to evaluate all candidates in an
objective manner. We’ve seen a distinct rise in the use and prioritization of
assessments in the last few years, and they’re now being leveraged
increasingly not just for hiring but also for promotions and internal mobility
opportunities. Some of the companies I’ve spoken with in recent months
include DotIn, Cangrade and Plum.
The second most common approach was auditing processes to ensure
fairness. One way to do this is to look at your company’s hiring funnel and
compute the number of diversity candidates at each stage of the journey. If
you’re getting in a good mix at the top of the funnel but only 20 per cent of
190 ARTIFICIAL INTELLIGENCE FOR HR

diversity candidates make it to the interview stage, you need to re-evaluate

your screening process. Using those types of measurements can help to
illuminate where opportunities exist to balance the scales.
The third option is also becoming an increasingly common practice:
resume blinding. Some of the providers offering this include Oleeo,
Eightfold and Talvista, a dedicated diversity hiring platform. The process
involves removing a person’s name and/or identifying characteristics from
the resume to ensure a fairer assessment. If the algorithm takes into
account any of the following seemingly innocuous items, they could be a
direct or indirect indicator of someone’s diversity status:
●●
name: can indicate gender
●●
university: can indicate gender or race
●●
postcode/zip code: can indicate socioeconomic status, a proxy for race
●●
activities: can indicate gender or race

It’s mind-boggling when we start to realize just how much an algorithm can
see and infer about a person, but that’s the importance of having a human
in the loop during decision-making. While some of the methods that are
available today may unintentionally cause more bias, there are others that
have these types of capabilities, allowing employers to ensure a fair and
equitable hiring process.

Beyond fairness, one other noteworthy component of this conversa-

tion is perceived fairness. Another study builds on this idea by showing
that humans in general believe that algorithms are less fair when it
comes to employment decisions because of the concept of reduction-
ism.6 In other words, we as humans think that algorithms will boil
down a decision to some simple factor while assuming a human deci-
sion maker can take in more inputs before making a determination.
This has been proven false on numerous occasions, including the
experiment on employee feedback analysis by an algorithm cited
earlier in Chapter 3. It’s been shown that algorithms can take into
account virtually limitless amounts of data to make a decision, whereas
humans can take in only a small fraction of that. Can algorithms care?
Show concern? Feel empathy? No, no and no. Can they make a deter-
mination based on thousands of data points? Absolutely.
 CHALLENGES OF ADOPTING AI TECHNOLOGY 191

Even if an algorithm is making determinations only on acceptable,

unbiased factors, if it feels like a mysterious, black-box decision, then
it may still be challenged by those affected by its determinations. In
one 2020 study performed by Modern Hire, a developer of science-
based hiring and screening solutions, the company found that the
specific disclaimer offered to candidates during the application
process can affect how they feel about the decision being made.7 This
suggests that clearer and more helpful descriptions of what factors or
signals the algorithm considers during the screening process can help
to alleviate fears for candidates applying for jobs. Interestingly, there
was no significant difference in reactions to AI screening based on
respondent race or ethnicity.

Technological complacency
If I threw out a random ten-digit number for you to remember, how
well do you think you would do? This used to happen regularly when
we would gather a phone number from a friend or relative, but today
we just plug that into our phones and move on. I have friends who
even forgot the phone numbers of their own spouses when they lost
access to their mobile device! In the next chapter I point out how
things like creativity must be nurtured and developed if we want to
maintain a human edge with our soft skills, but a more basic question
is whether AI handicaps our ability to think. If so, what can we do
about it, if anything? Kurt Marko, a writer for Diginomica, writes
about this in the context of human skills and how they atrophy over
time. He asserts that our continued reliance on automation will lead
to increasing losses in the skills we needed to accomplish those tasks
prior to automation. Instead of just relying on the systems until
human intervention is needed, they should be designed in such a way
that continuously reinforces our skill development and refinement.
Additionally, he points out that ‘convenience is the death of innova-
tion’ if we try to automate away opportunities for human creativity
and epiphany.8
192 ARTIFICIAL INTELLIGENCE FOR HR

In a conversation with Dale Kennedy, an HCM technology expert,

we discussed this concept in terms of self-driving vehicles. Kennedy
believes that AI has the capability to improve human resources
dramatically not just through automating tedious processes but also
by providing key information and insights that talent leaders need at
critical times. However, he explained that as vehicles get smarter and
smarter at driving autonomously, there is less and less need for human
interaction at the wheel. This can create critical weaknesses in our
ability to navigate situations and take over in times when the vehicle
runs into an unpredictable situation. Our senses and skills can become
dull without practice. Consider the example of the SourceCon
Grandmaster Challenge winner Randy Bailey profiled in Chapter 5.
He was able to beat the recruiting algorithm with a mix of tenacity
and ingenuity because his searching and critical thinking skills are
sharp and finely honed. If he was only using algorithms to automate
sourcing, searching and communicating with candidates, then he
would not have been able to beat the machine in my opinion.
In another example related to our topic, let’s imagine that your
firm begins using an algorithm to suggest pay rates for new offers
based on historical compensation data, market trends, company
performance, employee turnover data and other relevant inputs. The
system was implemented in hopes of levelling the playing field and
eliminating the pay gap between male and female workers. The tech-
nology seems to be working perfectly, but after a few years the system
begins suggesting higher and higher pay rates in job offers because it
sees that employee turnover drops as wages increase, even though the
business is not improving enough to cover the costs. If you were
running HR at this firm, how would you know when to take back
control? And, if you took control, would you know how to solve this
problem after several years of letting that skill lapse? In this scenario,
recruiters and/or hiring managers would have to be retrained on how
to negotiate salary with candidates, but the HR team would also
have to revitalize its knowledge and capabilities around compensa-
tion best practices, trends in pay transparency and other related
factors. A metaphor that makes this challenge easier to grasp is to
imagine your firm has a single person in charge of running all
 CHALLENGES OF ADOPTING AI TECHNOLOGY 193

­ ecisions around learning technology and content strategy. Then that

d
person leaves the firm without notice. How would you recover?
It really comes down to the fact that AI has the ability to make us
less intelligent, if we let it. In a conversation with the Head of Talent
for a public utility, she explained that her firm’s leadership team had
purchased a machine learning tool to help provide insight into a
particular business challenge. However, it quickly became apparent
that the business leaders weren’t using the system to help support
their decisions – they were using its recommendations as the decision.
There was no oversight, no input and no pushback on the recommen-
dations of the system, despite the fact that no algorithm is 100 per
cent accurate all of the time. She was frustrated at their inability to
see that the system was meant to be a decision support tool – a predic-
tive tool – not a decision-making tool. Her fear was that the team
would become too reliant on the system, fail to realize when it was
leading them astray, and be too far removed from the problem to
make any meaningful change to rectify the issue.
The core question I want you to consider is not just if we’ll realize
when we need to put our hands back ‘on the wheel’ in these situa-
tions. The bigger question is what we might be losing in the process.
The answer is to ensure that even in a process we’re automating, we
still have human oversight. We avoid black box algorithms that
compute and make predictions with no context or clarity around
how the decisions were made. We prioritize human skills like creativ-
ity, collaboration and critical thinking so that we don’t lose that edge
that allows us to overcome a wide variety of challenges in the work-
place, regardless of what they might be.

Algorithm aversion
Did you know that, according to Gallup data, more than nine out of
ten people who take a new job do so outside their own company?9 This
has always puzzled me, especially in larger companies, because there is
a plentiful supply of capable talent for employers to take advantage of,
yet they look outside for hiring decisions. When this is combined with
194 ARTIFICIAL INTELLIGENCE FOR HR

the data from the Wharton study we looked at earlier in the book that
shows external hires cost more and perform worse,10 it shines a glaring
light on one of the most common decisions we make. As organizations
and leaders, we’re more likely to look for an external candidate, in
part, because we don’t know all their flaws. We’ve never seem them
fail. We believe, for some reason, that a hire from outside the company
is going to perform better. This is a close parallel to the concept of
algorithm aversion we’re going to discuss below.
Algorithm aversion has a confounding effect on our hope to adopt
and leverage AI-powered software in the workplace, and it is similar
to the example of prioritizing external hiring. In essence, algorithm
aversion is the term used to describe our mistrust of the predictions
algorithms offer. In several studies, researchers have shown that we
are less likely to believe in algorithms than in human judgement, espe-
cially if we have seen them err.11 This is the same decision we make
when we avoid promoting our internal workers and instead seek an
outside candidate for a role. While the experiments were focused on
several different prediction scenarios, let’s apply this to the world of
HR that we live in. As you dig into some of the results and informa-
tion from the experiments below, look at them through the lens of
predicting the right candidate, helping a worker choose the right
development path, or finding the right engagement method for an
employee. Try to look at these findings through a common prediction
challenge you face, because it will make the concepts more tangible.
In one set of experiments, participants had to choose whether to
put their belief in an algorithm or in a human for a series of forecast-
ing activities. The researchers found that we as humans are much
more likely to believe in our own abilities over those of an algorithm,
and those differences became even more pronounced when a partici-
pant saw the algorithm make an error in its predictions. While we
might give our co-workers or even ourselves a break when we make
a poor decision, we’re less likely to give an algorithm a break when
that happens. The reason this is significant in the context of the work-
place is that algorithms will increasingly be used to support decisions
and even make decisions in some instances, and if there is a funda-
mental distrust of the information the algorithms suggest, then
 CHALLENGES OF ADOPTING AI TECHNOLOGY 195

adoption and usage will not go as planned. Additionally, let’s say an

algorithm is being used to help with removing bias from hiring deci-
sions. If that algorithm is avoided because the hiring manager doesn’t
think it’s making good suggestions, then the biased hiring processes
of old will continue to proliferate.
To get a more nuanced view of the specific issues that humans have
with their algorithmic counterparts, participants in the studies were
asked to rate the performance of humans against that of algorithms
on a specific group of attributes. The responses offer an incredibly
nuanced perspective of how we perceive ourselves relative to soft-
ware models. For example, participants said that algorithms were
better at avoiding obvious mistakes and weighting information in a
consistent manner. However, those in the experiments rated them-
selves as superior in detecting exceptions, finding the ‘hidden gems’
the software overlooks and learning from mistakes. Across the four
experiments the results of these measurements were sometimes less
pronounced, but one specific characteristic came out in favour of
human judgement every time: getting better with practice. So, what
does this say about us as humans? We give computers credit for things
that are obvious: they are unbiased and more likely to be consistent,
and that consistency can mean they don’t make those silly or obvious
mistakes that humans might by accident or when tired. On the other
hand, we wildly overvalue characteristics in ourselves. We think we
can detect exceptions in data, even though computers are much better
at detecting exceptions in large volumes of data. We think we learn
from past mistakes, but how many times have you made a mistake
more than once before you realized the root cause of the mistake?
And, finally, we believe we get better with practice. While that may be
the case, we also need to acknowledge the fact that computers can get
progressively better with practice and they do not share some of our
needs for things like rest.
But one of the conclusions by the authors at the end of the study is
that it didn’t explore deep enough into what it would take for humans
to believe in an algorithm. Yes, we know that people are less likely to
believe in an algorithm than in humans, especially after we’ve seen an
algorithm make a mistake. But how could we shift this perception so
196 ARTIFICIAL INTELLIGENCE FOR HR

that users actually trust what algorithms offer? In a later exploration

of this concept, one of the authors took on this challenge with inter-
esting results. Dietvorst built upon the original study by allowing
participants in the experiments to use algorithmic models to make
predictions, but then those participants were allowed to modify the
algorithms to a slight degree.12 The results are fascinating. When
those in the experiment had an opportunity to modify the algorithm,
they were more likely to be satisfied with the predictions of the algo-
rithm and, accordingly, more likely to use the algorithm in the future.
But the interesting part is the degree to which those modifications
could occur. Even if those modifications were insignificant, partici-
pants were more likely to believe in the algorithms. To explain,
consider the following scenario.
Let’s say you want to predict who is the right succession candidate
for an executive role. To begin, your algorithm presents a slate of
candidates. As we’ve already seen, you have an innate distrust of
algorithms, but as with the above experiment you have an opportu-
nity to modify the algorithm to some degree by making sure it takes
into account the factors you think are important to making the right
selection decision. Even if the inclusion of those factors doesn’t
change the slate of candidates one bit, you’re still more likely to
believe in the predictions it makes. I liken it to the example of when
someone has to speak up in a meeting to ‘say their piece’. Even if their
statements don’t fundamentally alter the direction of the decision
being made, the person feels better knowing they had their opportu-
nity to weigh in on the topic.
This idea of using modifiable algorithms that accept user inputs
(even if it doesn’t change the outcome) is one of the answers to solving
this problem of algorithm aversion. Another is to openly examine the
performance of the algorithm, comparing and contrasting with the
human decisions. The more we can shine a light on this bias, the more
likely we can overcome it. We know that algorithms and software can
offer better forecasts than humans, yet we have a tendency to discount
the value of those predictions in favour of our ‘gut’ or ‘instinct’. In
areas like predicting medical issues, this scepticism could be costly. In
terms of hiring or promotion decisions, it could be the difference
 CHALLENGES OF ADOPTING AI TECHNOLOGY 197

between success or failure. Or, in some cases, this ‘gut feel’ approach
could lead to less diversity and higher risk of litigation. Overall, it
demonstrates a key hurdle between where we are today and where we
want to go as an employer population. If we want the benefits of AI
technology like rapid processing, unbiased recommendations, and so
on, then we need to be willing to trust the recommendations and
predictions those technologies can offer. Otherwise, it will be like
running a race with your ankles tied together: you might eventually
get to the finish, but it won’t be fast and it won’t be pretty.
These findings could signify a few things that we need to be aware of.
First of all, we don’t really trust algorithms very much. Over time, as we
see the recommendations are relatively accurate, that trust will grow,
just like a relationship with a new co-worker blossoms as we build
trust. That said, the idea that we need to be heard reinforces yet again
the concept from the Uber drivers’ rebellion against Uber’s algorithm.
When employers rely too heavily on algorithms to make choices, the
people who are accustomed to being listened to are going to feel left
out. Imagine you are the expert on a certain topic in your workplace
and your employer buys a new system that automates recommenda-
tions that you normally provide. Even if this doesn’t displace you from
your position, it’s bound to put some measure of strain on you as your
voice becomes less and less ‘heard’ relative to the algorithm.
Because we as HR leaders have insight into the HR operations at
our firms, we have the unique opportunity to see these adoptions as
they cut across the employee population. We can look for ways to
ensure the people are still a part of the decisions to whatever degree
is necessary. Plus, it would be wise to think through what will happen
in the inevitable instance where the human worker disagrees with
something that an algorithm predicts. Who do you side with? How
do you do it in such a way that you don’t completely disengage the
worker, dismantling the value they’ve built around their role and
recommendations over time? Continued research in this field will
help to shed light on how we can work with our AI-based systems
and not against them, but it’s also up to us to help lay a foundation
of trust by adopting and utilizing these types of systems wherever
they can add value.
198 ARTIFICIAL INTELLIGENCE FOR HR

Data privacy and AI

One variable in the equation that may encourage workers to be less
trusting in artificially intelligent systems is data privacy. Data from
the Sierra Cedar HR Systems Survey shows a statistically significant
correlation between employers with a data risk and information
security strategy that includes HR systems and those employers that
have an interest in machine learning tools.13 This makes sense, as
employers with a keen eye for information security realize the treas-
ure trove of data that resides in the HR information systems. The
information in those systems requires protection, but has your firm
even thought about how it would answer the common questions
around this data, how it’s collected, whether or not it’s shared with
other entities, and what it’s used for?
Several years ago, I was speaking on a panel at a conference. The
topic of the session was HR analytics, and the room was packed
with business executives from firms around the globe trying to
understand how to adopt analytics in a way that fitted their unique
needs. A woman stood up during the question and answer portion
and asked the panel how she could take advantage of analytics when
her employees wouldn’t allow it. Intrigued, I asked her to elaborate.
It turned out that her firm was based in Germany, and the worker
council had a firm grip on what data was being collected from
employees and who could access it. In this example she had led a
project to gather data in an employee engagement survey with the
goal of improving the conditions and work environment for the
employees, but when the firm tried to bring in its third-party consult-
ing partner to analyse the data, things got tricky. The council would
not grant access to the third party to analyse the data on the firm’s
behalf. Because of that decision, the firm would either have to hire a
new employee on the HR team just to help work through the survey
data or shelve the whole thing, wasting the funds invested in the
development and capture of the survey.
At the time, I’ll admit this example seemed to be a little ‘out there’
for me. I remember wondering if employees really cared that much
about who saw their data and what it might be used for. Yet times
 CHALLENGES OF ADOPTING AI TECHNOLOGY 199

have changed. More and more public data breaches have caused
people to be more careful about who has their information and what
they can do with it. For example, in 2017 a data breach at Equifax
led to the exposure of personal information for 143 million people.14
Additionally, one online source points out that of the nearly 10 tril-
lion records lost or stolen in data breaches since 2013, only 4 per cent
of them were encrypted to render the data useless to hackers.15
If we add new laws like the General Data Protection Regulation
(GDPR) enacted in the EU in 2018, this landscape is definitely chang-
ing. GDPR rocked the HR world for months in advance of its adoption
as talent acquisition and HR leaders looked for information on how
to prepare for the law’s implementation and requirements for data
protection, and even today the aftershocks are still being felt in terms
of demand for training and information about what GDPR really
means for global employers and the data they collect and use. For
example, employers using recruiting technologies that store informa-
tion on applicants have much tighter restrictions on the data.
Employers need to be very clear around the areas of consent, disclo-
sure and retention of data.16 Another interesting point in the GDPR
law includes a requirement to provide the right to an explanation.
That’s right – consumers now have a legally protected right to ask for
an explanation of the algorithms that impact their lives.17 On a more
regional level, California has passed its own GDPR-like law, the
California Consumer Privacy Act, and other governing bodies have
looked to regulate AI, in particular within the US. For instance, Illinois
passed the Artificial Intelligence Video Interview Act to govern how AI
could be used for hiring.18 This law requires employers using AI
screening tools to notify applicants, explain what characteristics are
considered by the algorithm, gain their consent and delete the videos
within 30 days. New York City is considering a similar requirement
for employers to be transparent about using artificial intelligence
during screening processes. It also puts a burden on technology
providers to ensure their algorithms are not discriminatory.
So, while the original request from the German-based HR execu-
tive seemed to be out of place at the time, in hindsight it was a
foreshadowing of the concern around employee data privacy that
200 ARTIFICIAL INTELLIGENCE FOR HR

was to come. Despite it throwing me off a bit, the answer I gave to

that leader is still applicable today, and perhaps even more so. While
I didn’t have a simple answer for getting the council to change its
mind, I did encourage the HR executive to be very clear with the
employees about what kinds of data the company collected and how
it would be used. The example I gave her was around video stream-
ing. Regardless of the video streaming service you use, they all collect
data about your viewing patterns, preferences and more. The fact
that all of that information is being captured, stored and analysed
should worry us to some degree, right? After all, research in the 2016
Global Internet Phenomena Report shows that more than half of
downstream internet traffic for North America comes from just two
sources: Netflix (35 per cent) and YouTube (17 per cent).19 Yet we’re
not worried. Why? Because if our video streaming service knows us,
then it can offer us better content recommendations. We spend less
time browsing and more time enjoying. Or, if our favourite ecom-
merce website knows our buying patterns and preferences, then it
can help us keep our cupboards full of the things we like best. There
is a clear line of sight between the information gathered and the value
we receive back from those systems. The question we should all be
asking as HR and business leaders is what kind of value we’re offer-
ing back to our employees and how we can communicate that so they
understand and approve of the way we’re using their data. In order
to utilize AI to its fullest extent and make accurate predictions about
the future, we need to have access to comprehensive data sources.
This issue of data privacy is cropping up at just the right time because
it will force employers to be clear about the data they gather, how
they plan to use it and (if they’re smart) how that will benefit the
employees and candidates the data comes from.

System integration is lacking

I read several years ago about a company that flies over retail shop-
ping centres taking real-time photos of parking lots as a way to predict
 CHALLENGES OF ADOPTING AI TECHNOLOGY 201

future earnings. The more full the lot, the more the earnings should be,
on average. What if we could also integrate this information into
scheduling and staffing practices of retail stores so we knew on a more
granular level how many people to schedule for a particular shift and
when to ping contingent labour sources that could provide resources
quickly? The problem with that is a lack of integration. Many of the
ideas presented in this book hinge on the idea of a single source of
truth when it comes to data or a connection that enables disparate
systems to talk with each other. And I’m not talking about ‘upload a
spreadsheet file and we’ll update the system’ kinds of integration. I’m
talking about real-time or near real-time access to information from a
variety of systems, both inside and outside of HR.
These integrations form a sort of ‘connective tissue’ that enables
better decisions. The more information you have, the better the
prediction you can offer, right? And if we’re in a profession that lives
and dies by the predictions we make, then shouldn’t we want our
systems to have better connections? In recent years I’ve seen technol-
ogy firms take steps to open up their application programming
interfaces (APIs) and enable better connectivity for partners and
others, yet this is still not a strong capability for most vendors. APIs
are a way for technology providers to allow other systems to access
portions of the database, enabling data to flow in from one system to
the other, or even bi-directionally if designed for that purpose. The
problem? The artificial intelligence many vendors are offering is
making recommendations based only on data that lives in their
system because they don’t have the ability to read and access other
systems. That said, some vendors are trying to make headway in this
area by prioritizing this.
One great example is Checkr, the company that powers back-
ground screening for some of the world’s largest and fastest-moving
companies. Checkr’s team describes it as an ‘API-first’ company,
embedding the product directly into the hiring experience and pass-
ing the data into the applicant tracking systems or into the interface
for candidates.
Another is HiBob, an HR engagement and communications plat-
form. The system hooks into systems like Slack and Microsoft Teams,
202 ARTIFICIAL INTELLIGENCE FOR HR

allowing workers to engage with the system without ever having to

leave the tools where they spend the most time.
And, lastly, Cronofy also takes an API-led approach. This schedul-
ing system powers interview scheduling for some of the world’s
largest recruiting technology providers, such as Personio, Outmatch,
Symphony Talent, Phenom and more.
The takeaway for you is to talk with your own vendors and poten-
tial partners to let them know this is a priority for you. Imagine being
able to see performance, learning, survey feedback and compensation
data in one place so that you understand whether you’re paying your
managers in an equitable way. Or that you have your learning and
recruiting systems connected so that your internal employees can see
the open jobs and the training they could take to improve their
competitiveness for those open roles.
Another example to wrap up with here is chatbots. E W Scripps is
a 4,000-employee firm that’s been in business since 1878. In a recent
presentation by the firm’s head of human resources, he explained
how the firm had connected its chatbot, provided by Socrates.ai, into
a variety of HR systems so that employees could access virtually any
information for pay, leave, benefits or policies simply by conversing
with the bot. This goes beyond simply answering questions based on
a predefined list of common issues employees face and allows the
system to automate much more of the process.20
This lack of connectivity among business and HR systems is a
challenge. Some of the solutions in the market designed to bring data
sources together or to provide blockchain-based credentials in a
vendor-agnostic way provide hope that this data challenge has the
potential to be solved. It won’t be resolved overnight, but the good
news is that unlike some of the other challenges presented here that
are based on human behaviours and are difficult to change, this is a
technological issue. Like any technology problem, this can be solved
with the right attention and creativity.
In the end, the more mature the connections become, the better the
data will become. The better the data becomes, the greater the value
that AI can produce for businesses and the workforce.
 CHALLENGES OF ADOPTING AI TECHNOLOGY 203

CASE STUDY
Beware the trolls: A cautionary tale of releasing AI into the wild

On 23 March 2016, Tay entered the world. Tay was a bot designed by Microsoft
to mimic a 16-year-old girl’s personality and interaction style on the web. This
was to be accomplished by having Tay mimic the speech patterns of Twitter
users around the world interacting with the bot. The concept was interesting –
can a bot learn to be more human by interacting with humans? In the end, it
turns out that trolls (people who post inflammatory messages online for their
own entertainment) were the ones that sank Tay’s prospects for the future.
At the end of the experiment, Tay was repeating racist remarks and making
up original offensive content as well.21 One might imagine that this experiment
took place over the course of several months, with the software being corrupted
over time by the volume of negative interactions. In reality, it took less than 24
hours for the AI-powered bot to devolve from a cheerful teen into a hate-
spouting monster. Some people, and certainly the programming team from
Microsoft, would claim that this is not a fair example because users attempted to
‘hack’ the bot by intentionally providing leading questions, false information and
improper language in a stream that built up the system’s knowledge base and
awareness of the world. Because this occurred over such a short time, that
volume of (bad) information became the dominant portion of the system’s
programming.
For practical purposes, AI is essentially like an infant. It requires specialized
care and ‘feeding’, especially in the earliest days after it is released into the
world. There is great potential, but there is also great danger as the AI has no
awareness of right and wrong. Microsoft’s experiment with Tay is a great
example of how employers and other firms should not launch tools into public
forums without first considering the implications of bad actors interacting with
the system. Whether for humour or other more serious purposes, users fed Tay
so much negative information that it never really recovered.

Despite the challenges…

Despite the challenges outlined here, in previous chapters we’ve
explored the significant value that artificial intelligence technologies
can provide. More than ever before, I’m convinced of the paradoxical
204 ARTIFICIAL INTELLIGENCE FOR HR

nature of technology adoption. In order to adopt technology that can

help eliminate some of the human error, we need to take a very human
approach to change management, influencing others to get that tech-
nology in place. It’s perhaps even more applicable here in this
discussion of AI and algorithms as they are more likely to filter out
those human ‘issues’, like biases, around decision-making that other
technologies never really touch. Accountants have a general ledger
that tells the financial truth. Marketers have their customer relation-
ship management and marketing automation tools that highlight
prospect interests and activities. But HR software? We deal with the
grey areas of people decisions, which are usually much less clear-cut
than some of these other examples.
When we think about bringing artificial intelligence technologies
into the workplace, we might not run into issues like Microsoft did
with Tay. How many of us really need to be integrating a fake teen AI
into the rhythm of our workforce communications? However, we
will inevitably run into people that try to stretch the system and find
its loopholes. For instance, there are countless examples in forums
online where users have attempted to have off-script conversations
with chatbots, offering up screenshots of the interactions for the
comedic value. We intuitively want to find the edges, the holes and
the issues with technology as this helps us to prove that we’re more
human and ‘better’, by whatever subjective measure we’re using at
the time of the estimation.
During one interview with Cecile Leroux, an executive at Ultimate
Software, she explained that the company’s approach to AI develop-
ment was less about automation and more about augmentation. In
other words, it’s really about helping today’s HR leaders do great
things in better ways instead of just automating manual tasks, which
brings to mind images of robots replacing humans. In a profession
that is supposed to be the most human-centric function within an
organization, it’s worth remembering that sentiment as it can change
how people perceive the advance of AI and the value it can bring.
During the early 1800s, English Luddites revolted in an attempt to
protect their jobs and destroy automation machinery in the weaving
 CHALLENGES OF ADOPTING AI TECHNOLOGY 205

mills. While that sort of outcry is less likely today, it’s no less i­ mportant
to think about the perceptions of the workforce when more and more
of the workload is being offloaded to algorithms and computers.
Especially when those decisions are critically important, such as those
surrounding hiring, training, succession and more.
This goes back to the conversation about the ‘AI effect’ in Chapter 2.
Humans want to feel like they are different from or better in some
way than technology, so we make an effort to discount advances in
artificial intelligence technology by writing off improvements as
something less than intelligence. In other words, we keep moving the
bar that defines ‘intelligence’ every time an algorithm or computer
meets the bar. It’s that need to differentiate ourselves that pushes
some people to dig into a bot’s inner workings or feed it information
that may change its behaviours. We need to expect some level of this
when AI becomes a more integrated part of the workplace. The back-
lash, for lack of a better term, might not be as severe as we saw in the
Uber driver ‘rebellion’ example. Workers might not be so frustrated
and powerless that they seek out ways to fight back and sabotage the
algorithms that make your workplace function. However, it’s foolish
to expect complete acceptance by everyone. Additionally, it’s impor-
tant to remember that in the Age of AI (if I’m allowed to offer such a
grandiose moniker), the human component is still an essential piece
of the puzzle. As John Roese, the President and Chief Technical
Officer of Dell EMC, put it, ‘[Workers] understand that AI and
machine learning will eventually disappear into the infrastructure,
becoming second nature. They know that machines are non-human
and that human ingenuity is irreplaceable.’22 I have repeatedly stressed
and provided example after example within this book that this is not
an either/or solution, it’s an and solution. We need people and tech-
nology, and each has a distinct set of strengths that we can take
advantage of to solve problems and advance towards our goals as
business leaders. As we’ll explore in the next chapter, the ability for
us as humans to use our skills to remain relevant, competitive and
valuable will become increasingly important as the level of a­ utomation
increases in the workplace.
206 ARTIFICIAL INTELLIGENCE FOR HR

KEY POINTS
●●
Bias is a real and present danger in decision-making processes, and
while bias has always been an issue, we need to be careful not to hand
off our decisions to an algorithm without clearly understanding how
predictions and recommendations are made and whether bias is
inherent or not.
●●
As humans, we want to trust our instincts, even if they’ve been shown
to be incorrect in the past. The concept of algorithm aversion may slow
adoption of automated technologies, but adoption will certainly increase
as we define clear links between the outcomes and the value these tools
can provide.
●●
Data privacy and security have never been more of an issue than in
today’s workplace. With more data than ever before being created,
stored and accessed, employers are creating liability for themselves.
Understanding legal requirements and how they apply is a good first
step. Leaders should also prioritize these components for internal
technology teams and for software partners.

Notes
1 M Hornbeck. This machine read 3.5 million books then told us what it
thought about men and women, World Economic Forum, 30 September 2019.
www.weforum.org/agenda/2019/09/men-women-books (archived at https://
perma.cc/LH7H-RUEF)
2 Y Shoham, E Brynjolfsson, C LeGassick, R Perrault, J Clark and J Manyika.
The 2017 AI Index Report, Stanford University, 2017. https://hai.stanford.edu/
ai-index-2017 (archived at https://perma.cc/Y7DE-FB3N)
3 L Sweeney. Discrimination in online ad delivery, 28 January 2013. https://
papers.ssrn.com/sol3/papers.cfm?abstract_id=2208240 (archived at https://
perma.cc/652R-6C9P)
4 K Johnson. Pymetrics open-sources audit AI, an algorithm bias detection tool,
Venture Beat, 31 May 2018. https://venturebeat.com/2018/05/31/pymetrics-
open-sources-audit-ai-an-algorithm-bias-detection-tool/ (archived at https://
perma.cc/57J3-EG9P)
5 B Eubanks. We asked 800+ employers about virtual recruiting, Lighthouse
Research & Advisory, 6 April 2021. https://lhra.io/blog/report-we-asked-800-
employers-about-virtual-recruiting-new-research/ (archived at https://perma.
cc/5NG4-NUWG)
 CHALLENGES OF ADOPTING AI TECHNOLOGY 207

6 DT Newman, NJ Fast and DJ Harmon. When eliminating bias isn’t fair:

Algorithmic reductionism and procedural justice in human resource decisions,
Organizational Behavior and Human Decision Processes, 2020, 160,
September. www.sciencedirect.com/science/article/pii/S0749597818303595
(archived at https://perma.cc/THN8-7ZEK)
7 A Zullo. Modern Hire reveals first-of-its-kind research on AI consent language in
hiring, PR Newswire, 12 November 2020. www.prnewswire.com/news-releases/
modern-hire-reveals-first-of-its-kind-research-on-ai-consent-language-in-
hiring-301171385.html?tc=eml_cleartime (archived at https://perma.cc/
DLW7-MRMD)
8 K Marko. Beware the AI risks of over-automation and hyper convenience,
Diginomica, 20 February 2018. https://diginomica.com/2018/02/21/beware-
the-ai-risks-of-over-automation-and-hyper-convenience/ (archived at https://
perma.cc/74EW-5FAV)
9 B Rigoni and B Nelson. When making career moves, Americans switch
companies, Gallup, 22 October 2015. https://news.gallup.com/
businessjournal/186311/making-career-moves-americans-switch-companies.
aspx (archived at https://perma.cc/3XPY-SP8W)
10 M Bidwell. Paying More to Get Less: The effects of external hiring versus
internal mobility, University of Pennsylvania, 2012. https://faculty.wharton.
upenn.edu/wp-content/uploads/2012/03/Paying_More_ASQ_edits_FINAL.pdf
(archived at https://perma.cc/B5MA-DE8T)
11 B Dietvorst, J Simmons and C Massey. Algorithm aversion: People erroneously
avoid algorithms after seeing them err, Journal of Experimental Psychology:
General, 2015, 144 (1), 114–26. https://repository.upenn.edu/cgi/viewcontent.
cgi?article=1392&context=fnce_papers (archived at https://perma.cc/X3XF-
5Q3V)
12 B Dietvorst, J Simmons and C Massey. Overcoming algorithm aversion: People
will use imperfect algorithms if they can (even slightly) modify them, 5 April
2016. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2616787 (archived
at https://perma.cc/V92F-NFCT)
13 Sierra Cedar. HR Systems Survey white paper, 2021. https://sapientinsightscom.
sharepoint.com/sites/WebDownloads/Shared%20Documents/General/Sapient_
Digital%20Report%202020_Final%2010-26-2020.pdf (archived at https://
perma.cc/D2UB-G385)
14 Federal Trade Commission. Equifax data breach settlement, January 2020.
www.ftc.gov/equifax-data-breach (archived at https://perma.cc/D2UB-G385)
15 Thales Group. Data breaches compromised 3.3 billion records in the first half
of 2018, 23 October 2018. www.thalesgroup.com/en/markets/digital-identity-
and-security/press-release/data-breaches-compromised-3-3-billion-records-in-
first-half-of-2018 (archived at https://perma.cc/M746-ZV7Y)
208 ARTIFICIAL INTELLIGENCE FOR HR

16 R Maurer. New EU data law will change how you engage with job applicants,
SHRM, 28 February 2018. www.shrm.org/resourcesandtools/hr-topics/
talent-acquisition/pages/gdpr-eu-data-law-hr-recruiting-shrm.aspx (archived at
https://perma.cc/VNJ3-4RAA)
17 S Amornvivat. Is it possible to be too data driven? Bangkok Post, 21 February
2018. www.bangkokpost.com/opinion/opinion/1415298/is-it-possible-to-be-
too-data-driven- (archived at https://perma.cc/928Y-8LWC)
18 Davis Wright Tremaine LLP. Employers using AI in hiring take note: Illinois’
Artificial Intelligence Video Interview Act is now in effect, JDSUPRA, 11
February 2020. www.jdsupra.com/legalnews/employers-using-ai-in-hiring-
take-note-54767 (archived at https://perma.cc/7D8H-6UWW)
19 Sandvine. 2016 Global Internet Phenomena, Sandvine, 2016. www.sandvine.
com/hubfs/downloads/archive/2016-global-internet-phenomena-report-latin-
america-and-north-america.pdf (archived at https://perma.cc/69X9-FWPS)
20 LeapGen. How AI can help transform the workforce experience at the E.W.
Scripps Company, LeapGen YouTube channel [video], 2018. www.youtube.
com/watch?v=1H8X3Ac9fgQ&t=14s (archived at https://perma.cc/45LJ-
JDG6)
21 J West. Microsoft’s disastrous Tay experiment shows the hidden dangers of AI,
Quartz, 2 April 2016. https://qz.com/653084/microsofts-disastrous-tay-
experiment-shows-the-hidden-dangers-of-ai/ (archived at https://perma.cc/
J6KW-G3D9)
22 R Banham. Easing in the robots: How to confront employee fears, Dell
Technologies, 28 March 2018. www.delltechnologies.com/en-us/perspectives/
easing-in-the-robots-how-to-confront-employee-fears/ (archived at https://
perma.cc/CEM2-F89V)
 209

09

HR skills of the future

When a pathologist needs to diagnose cancer, it requires painstaking

scrutiny of biopsy samples to identify cancerous cells. This is the
proverbial search for a needle in a haystack, since there could be
millions of healthy cells and just a few cancerous ones in any given
sample, if cancer is even present in the sample at all. A series of exper-
iments unveiled by Harvard in 2016 involving the analysis of breast
cancer tissue cells showed us a stunning set of results.1 Teams were
assigned the task of identifying metastatic cancer cells, because breast
cancer is one of the most common types of cancer in addition to
being one of the most deadly. The task itself holds significance beyond
simple identification, because these types of cells are typically gath-
ered in a biopsy where cancer is suspected but not yet confirmed.
Therefore, getting better results at this stage of the process can
improve life expectancy for those who have breast cancer.
Within the context of the experiment, the team’s artificial intelli-
gence algorithm spotted cancer with 92 per cent accuracy. However,
when compared with human pathologists, the AI fell slightly behind.
Human experts were able to spot the cancer cells with 96 per cent
accuracy. But this isn’t another ‘us versus them’ story of humans and
machines. The real success story happened when the team partnered
up both resources to answer the question: are we better together?
The answer was clear – in the partner experiments, blending the algo-
rithm with human judgement led to identification of more than 99
per cent of cancerous biopsies. The repercussions are critical: for that
3 per cent of women who would have been misdiagnosed by human
pathologists, this could give them a fighting chance.
210 ARTIFICIAL INTELLIGENCE FOR HR

This type of research looking at humans and machines has been

carried out in other fields as well. Take chess, for instance. When
humans are paired with computers (called a ‘centaur’ after the myth-
ical half-man, half-horse creature) during games, they can perform
better than computers or humans by themselves. That’s because
humans are great at broad strategy and direction while computers
are very good at rapid calculation and deeply validating the next
activity. It’s not ‘either/or’, it’s ‘and’.
What do these types of experiments mean? Beyond the obvious
answer that humans and machines complement each other’s capabili-
ties, this points to something deeper. This kind of finding highlights
the fact that what we need from a doctor may be changing. In this
example, it’s possible that AI would not be able to identify more rare
forms of cancer or would not be able to compute an erroneous image,
potentially missing a critical diagnosis. In those cases the images could
be flagged for human intervention. While they still need to be able to
verify a computer’s output, doctors will spend less time doing that
type of work and more time doing other tasks, many of which could
be patient-facing, which may shift the type of skills necessary for
success in the medical field. With all due respect to the lovable health-
care robot Baymax from Disney’s Big Hero 6, if I was going to receive
a diagnosis of a serious health issue, I’d want to get that from a living,
breathing human being, not an unfeeling machine. For some tasks,
human intervention is not only a good idea in general, it’s demanded
by the population being served. This shift is inherently similar to the
one facing the HR profession and the business more broadly.
In one study, we asked workers what skill matters most to long-
term career success. Without prompting them with any options, the
most popular answer was communication skills. The same question,
asked of HR and talent professionals, led to the same answer as well.
This was true before 2020, but in a world that is increasingly remote
and virtual one could argue that it’s doubly important today. In our
research, we’ve found that the skills needed for HR and talent profes-
sionals in a more automated world are changing, and while hard
skills matter, it’s actually a blend of both hard and soft skills that is
driving performance. Things like analytical skills help to drive return
 HR SKILLS OF THE FUTURE 211

on investment and build a business case, but capabilities like commu-

nication and collaboration lead to innovation and engagement.
Broadly, the market has tended to emphasize soft skills in recent
years. Doniel Sutton, who has experience as Global Head of People for
multiple Fortune 500 organizations, believes that out of all the skills
that the workforce can have, agility is the most important.2 Regardless
of the role, Sutton says that agility is the single most critical skill that
helps workers to stand out from the crowd and succeed. From bounc-
ing back against ever-present change to learning on the fly, agility
encapsulates much of what makes workers successful today. You might
imagine the head of hiring and development decisions for a technology
firm would say that some coding language or development methodol-
ogy was the biggest factor, but instead it’s a soft skill. Therefore, in a
world of increasing automation, what sort of skills are necessary to
drive the future of work? How will the demand for hard or soft skills
fluctuate based on the increasing capabilities of bots, algorithms and
machines? Will soft skills become the currency of the future?
There are two ways to approach these types of questions. The first
is to ask what types of things humans can do that machines can’t, at
least at this current stage of development. However, there are stories
almost daily about how these kinds of barriers are being broken
down by new robots and tools being developed across the globe. For
example, Boston Dynamics’ robot can open doors.3 Additionally, the
Henn na Hotel in Japan is almost entirely staffed by robots.4 A better
question might be to consider how humans can augment what
machines can do. This is interesting in itself because we’ve long seen
the equation from the other perspective: how can machines augment
what humans can do? Shifting this perspective can guide us towards
the right skills that matter not just today but also into the future, as
more and more of what we do is able to be automated.

The spectrum of augmenting humans with AI

If you wanted to skip the rest of this chapter, here’s the gist: artificial
intelligence will change the work HR does, but it won’t change the
need for the ‘human’ in human resources.
212 ARTIFICIAL INTELLIGENCE FOR HR

It’s my take that this change is driven partly by automation and

partly by augmentation, filling in gaps in what humans can do with
the complementary capabilities of algorithms. An interesting model
established by Boston Consulting Group takes this conversation
around augmentation even deeper to a more practical level.5 The
model lays out five specific varieties of human–AI interaction as well
as how to identify them situationally:
●●
Automator: Best for simple context and fast decisions.
●●
Decider: Best for situations where AI has context but humans need
to carry out the decision.
●●
Recommender: Best for multiple, repetitive decisions where AI can
make recommendations for humans to choose.
●●
Illuminator: Best for creative work where humans can benefit from
AI-driven insights.
●●
Evaluator: Best for high-stakes, unpredictable situations where
humans lead and get input from AI.

This approach is very similar to the one we’ve taken with the
Lighthouse Research & Advisory Human AI Decision Support Tool.
It offers a simple framework for making decisions about when and
where to apply AI, along with examples to help determine what
humans should be doing and where AI can play a leading or support-
ing role. This may change over time as the tools become smarter and
more sophisticated, but there will always be tasks that require a
human component.
In reality, there’s no true way to know just how much automation
will change the workplace (Figure 9.1). Estimates are just that –
educated guesses. Earlier in the book I pointed out the search engine
optimization example of how jobs change and new ones are created
that we couldn’t even foresee. However, let’s suspend disbelief for a
bit and explore some of the research that exists, exploring the big
picture thinking behind just how much automation could potentially
impact the workplace. If it helps, consider these numbers in the
context of direct impact on your entire HR team, if not the whole
organization. For example, if a study says to expect 40 per cent auto-
 HR SKILLS OF THE FUTURE 213

FIGURE 9.1 Human vs. AI task decision tool

Non-routine Executive
Employee coaching, change
surveys, career management,
exploration culture-building

Human led, Human led,

AI delivered human delivered

AI led, AI led,
AI delivered human delivered
Payroll
exceptions, Interviews, pay
benefits conversations,
reconciliation, talent reviews
Routine resume screening

Process focused Human focused

mation, consider what that means in the context of your HR function.

Would you expect to see 40 per cent fewer people? Have 40 per cent
more impact on the business? What would that translate into in terms
of creating a more personalized or high-touch employee experience?
Mahe Bayireddi, CEO and co-founder of Phenom, points out a
critical point: automation in itself doesn’t create new value. It’s in
how we use that time that we create value for our organizations and
stakeholders. If you save two hours a week and spend it on social
media or another unproductive activity, what is the overall benefit?
But if you save two hours a week and reinvest that time into better
relationships with stakeholders, understanding the latest technolo-
gies or honing your data analysis skills, that automation is truly a
value-add for you and your team. Bayireddi agrees: ‘HR leaders use
AI to create unforgettable talent experiences while using their
newfound time to forge more meaningful relationships will be rivet-
ing and truly transformative to the industry.’
With regard to automation predictions, McKinsey reports that
nearly half of all work could be automated by 2055.6 This prediction
is based on a model that looks at the skills of a particular role and
how predictable the role may be. McKinsey’s team shows that for
jobs requiring physical and manual skills or the most basic cognitive
abilities, the demand will shrink considerably by as soon as 2030. On
214 ARTIFICIAL INTELLIGENCE FOR HR

the other end of the spectrum, social and emotional skills are predicted
to rise in importance over the next decade. One of the key factors in
these types of research is the predictability of a job. Some parts of HR
are highly predictable (payroll, recruiting processes, etc.). However,
other parts are more complex (coaching leaders, designing employee
experiences, and so on). The ability to deal with unpredictability is
actually a powerful human skill. The concept of learning environ-
ments, both kind and unkind, were touched on earlier in this book.
Humans who are more generalized in nature have a better chance of
surviving and dealing with unpredictability, according to the research
laid out in the book Range.7 The author demonstrates that life is
more like an infinitely variable game with unwritten rules, little to no
feedback, and uncertain parameters. This, fundamentally, is an
unkind learning environment because it makes it difficult to learn
from our decisions in a quick and decisive way. An activity like golf
or chess offers a kind learning environment: clear rules and quick
feedback in the form of a poor shot or a lost game if we don’t perform
adequately. Algorithms are designed to be specialized, so they may
consume pieces of the work HR does, but they will not quickly over-
take the activities that are more generalist in nature.
In fact, research shows that becoming too specialized can actually
have negative impacts on our performance. For instance, students
that get too specialized in their own career/research field perform
worse than random on logic puzzles. If you become too rules-oriented
and constrained, you try to see every problem through the myopic
lens of your craft. In layperson’s terms: if you’re a hammer, every
problem looks like a nail. The issue is that this creates scenarios
where the things you learn in your craft may actually make you worse
at solving problems instead of better. A more diverse set of perspec-
tives can help individuals solve problems better, whether at home or
at work. That’s a powerful call for diversity overall, but also for
expanding our own skills as HR practitioners to be able to adapt and
change as our organizations evolve.
In an analysis more closely related to HR, one European firm
surveyed HR practitioners to understand their own perceptions of
what AI might mean for the profession.8 The report pointed out that
 HR SKILLS OF THE FUTURE 215

employers will be looking for different types of competencies from

candidates as time progresses, because different skills will be neces-
sary in a world with greater automation. Those skills include the
ability to solve complex, challenging problems, a keen focus on
service, critical thinking and people management. Respondents in the
survey were almost evenly split on several fronts, including whether
AI would lead to increases or decreases in jobs, whether HR’s role in
the business would grow or diminish and whether AI was a threat or
an opportunity.
Additionally, in a study of nearly 2,000 experts in technology and
market trends, the Pew Research Center found that this group was
evenly split on whether the advances in artificial intelligence would
eliminate jobs or not.9 In the study, key leaders at a variety of tech-
nology firms pointed out the opportunity for job creation with the
new robots, algorithms and learning systems in place, explaining that
these new jobs might be drastically different than what has existed in
the past. If I can put on my futurist cap for a moment, I believe this
job creation can also apply to the HR profession. For instance, in the
learning and development profession, one of the newer roles for
forward-thinking employers is a learning experience designer. These
specialists design learning experiences that engage learners with
content formats, types and delivery mechanisms that create addi-
tional value beyond static, transactional methods. Just as the learning
and development profession now has individuals focused on learning
experience design, it’s possible that enterprise organizations will have
key positions focused on designing candidate and employee experi-
ences that create emotional and rational connections with every
individual. Building these experiences will require not only advanced
versions of the AI technologies we have today but also the human
intuition necessary to guide those technologies to successful outcomes.
The takeaway here is that whether the audience is a group of HR
professionals doing the work or a group of high-level thinkers and
academics with a broader view of the world, both groups are equally
puzzled about what the future might hold in this area. What matters
on a more fundamental level is the expectation and anticipation of
change and what it means for the workplace. Companies will change.
216 ARTIFICIAL INTELLIGENCE FOR HR

Jobs will change. People will be forced to change along with them. In
virtually every organization, HR is the one tasked with change
management. However, one thing I’ve learned over the course of my
career is that HR can often fail to see the impact that might be
happening within its own function. What skills will be unnecessary,
or at the very least deprioritized, over the course of time?
Consider the following example of human performance that leaves
us awestruck. In 2006, Alex Haraguchi sat in a public location with his
eyes closed, talking. And talking. And talking. For more than sixteen
hours, Haraguchi recited 100,000 digits of pi from memory. For most
of us that took mathematics, rounding pi to 3.14 was good enough to
get the job done, but this Japanese man has dedicated a significant
portion of his life to memorizing digits of pi. However, a computer
could easily list pi digits beyond 100,000 without requiring years of
training and sixteen-plus hours of vocalization.10 One of my new
standby jokes when it comes to this discussion of shifting skills needs
in the HR function is this: ‘You know that person on your HR team
that’s really detail-oriented? She might want to learn another skill.’ As
we’ve seen in numerous examples, even the most detail-oriented human
on the planet isn’t a match for an algorithm. Machines can compute
faster, work longer and make fewer errors than humans when it comes
down to raw processing tasks. I know, I know. This isn’t a surprise to
you to hear that, but I will challenge you with this: how are you honing
your own skills to prepare for a more automated workplace? More
importantly, what skills should you be focusing on?

The importance of soft skills

When we consider the applications of AI technology and the oppor-
tunities for automating an endless supply of technical tasks, the skills
that will be most difficult to replicate are soft skills. Most adult
professionals have taken a soft skills course here or there, and some
have even had deeper dives into training and development of specific
skills around leadership or other competencies. Unlike hard or tech-
nical skills, which can usually be measured by throughput or work
quality, soft skills are more intangible or personality-related.
 HR SKILLS OF THE FUTURE 217

In some businesses and professions, individuals with strong techni-

cal skills can get away with a lack of soft skills. While some businesses
put more value on ‘how’ things get done than just looking at the end
result, many of them are willing to put up with highly productive
performers even if they don’t really get along well with others. From
an HR perspective it’s an incredibly common problem. But is this era
of strong technical skills and poor soft skills coming to an end?
In a discussion with Matt Chambers, CEO of recruiting technol-
ogy provider Loxo, one of his key perspectives is that recruiting
professionals shouldn’t be spending time on tasks that do not add
value. Instead, we should be automating those items and spending
more time on the higher-function tasks. By taking this approach,
organizations can automate the activities of high performers so low
to middle performers can increase their productivity. The question is
this: when that top performer’s job is automated to some degree and
it levels the playing field, will businesses put up with the lack of soft
skills any more? It stands to reason that the lack of those personality
characteristics will stand out much more when they aren’t hidden in
the shadow of top performance. This is just one more rallying cry
behind the idea that soft skills will become hard requirements. The
research our team completed in 2021 backs this up. As organizations
see more automation changing the talent acquisition space, they
agreed that a mixture of hard (analytics, technology) and soft
(communication and a consultative approach) skills would be most
important for upskilling the recruiting profession.11 Those intangible
elements including everything from teamwork and collaboration to
leadership, confidence and more will no longer be a ‘nice to have’ but
a ‘must have’ for the workforce.
Still not sure? Consider the types of work that will be left over if
the routine tasks are automated, whether in HR or without. Those
tasks will be high-touch, very personalized and strategic to some
degree. Can someone without the requisite soft skills pull that off or
will they flounder? As has often been said, the soft stuff is the hard
stuff. Soft skills lead to hard outcomes, and virtually any leader that
has led successful teams knows this is the case.
218 ARTIFICIAL INTELLIGENCE FOR HR

In a related discussion with Egan Cheung, former Vice President

of Product at Achievers, he explained his view of how jobs will
change with increasing automation. Specifically, he believes that as
AI and other tools change the talent and job ecosystem, we’ll see
new jobs that are more focused on the things humans excel at. In
other words, we’re not just seeing different jobs, we’re seeing more
humanized jobs emerge from this trend. As a self-described technol-
ogy optimist, Cheung wants to see technologies that are highly
personalized to help augment the capabilities of managers in the
workplace. A significant portion of employee satisfaction and
engagement at work is based solely on the individual’s direct
manager – just think about times you worked for a really great (or
really terrible) boss and you know exactly what I mean. Managers
need to be guiding their people with individualized care and atten-
tion, not just a ‘peanut butter spread’ approach to management
where all employees get the same treatment. But, to be honest, that’s
very challenging to pull off as a busy leader with other tasks and
responsibilities, which is why it often falls off the map, if it’s even
considered at all. Cheung says we should get to a point where
systems can see what employees do well and what they need to
improve, and then managers can get tailored advice and insights
into how to improve employee engagement and performance. For
those managers that don’t have the requisite skills to manage well,
algorithms can help bring them up to par. For those great managers
that already do a good job of this high-touch approach, technology
can help to streamline and refine their style.
If we break down jobs into their requisite tasks, the more predict-
able components will be more easily automated. Research from
McKinsey shows that more unpredictable work, or work with a high
degree of soft skills, will be harder to automate than other types of
labour.12 Figure 9.2 demonstrates the range of activities in the aver-
age job. The farther down the list you go, the harder it is to automate.
Replacing predictable physical activities has already happened on
some assembly lines with robots, but replacing the person that
manages others isn’t so easy, as we observed in Chapter 7.
FIGURE 9.2 Breakdown of average job tasks

Predictable physical tasks

Data processing

Data collection
Tasks become
increasingly
Unpredictable physical work
difficult to
automate as
Stakeholder interactions you move down
the spectrum

Applying expertise

Managing others

0% 2% 4% 6% 8% 10% 12% 14% 16% 18% 20%

SOURCE Adapted from McKinsey and Company

219
220 ARTIFICIAL INTELLIGENCE FOR HR

The rest of this chapter is dedicated to an examination of five

fundamentally human skills that will become more and more impor-
tant over time, regardless of the job level, function or industry. While
important within HR specifically as an employee-focused function,
they also have broad applicability for executives, managers and line
employees across the globe. You might note that there are times that
these skills overlap or complement each other in some areas, which is
to be expected. If you’re familiar with the story of Benjamin Franklin,
one of his well-known habits was cycling through a list of thirteen
virtues on a regular basis to improve his faculties and performance in
all areas of his life. While I can’t speak to the efficacy of his approach,
that type of repetitive, intentional practice with the following list of
soft skills might be the best way to start down the path to honing
your own.
Those skills, in no particular order, are creativity, curiosity, compas-
sion, collaboration and critical thinking. All of us have some of these
capabilities in greater measure than others, but all of us have at least
some capacity to grow and develop in each specific area. For instance,
creativity is one of the skills that either comes naturally for me or is
spurred more readily by the types of information I consume and the
white space I create for thinking. Whatever the case, it’s a compe-
tency I can readily leverage at virtually any point in time. At the same
time, compassion is one that I have to work harder at, because I am
hard-driving and sometimes lose sight of the fact that others need my
support or affirmation. That one requires more intentional effort on
my part to develop. It’s often been said that soft skills are easy to
define and difficult to cultivate. If you walk away from this book
with one overwhelming message about skills and competencies, it’s
my hope that you see the soft skills listed in Figure 9.3 as an oppor-
tunity to future-proof yourself against disruption from automation
technologies.
It’s important to point out that the model in Figure 9.3 was
designed using a mixture of interviews with business leaders and
academics both inside and outside the HR function. Additionally,
that research was combined with a thorough meta-analysis of a vari-
ety of research reports and data sources focusing on the future of the
 HR SKILLS OF THE FUTURE 221

FIGURE 9.3 Skills of the future

workforce. For example, Dr Vivienne Ming, a neuroscientist who

explores the intersection of AI and child development, believes that
social skills will take on increasingly important roles in a more auto-
mated world. Ming says that as jobs become more automated, even
low-level roles will require more creativity and collaboration.13 The
four categories she believes will become focal points are social skills,
metacognition, creativity and personality (motivation, purpose and
mindset). These points line up with a variety of other research efforts
and insights from scientists, notable business leaders and industry
experts. For instance, Chief Learning Officer magazine shared the
data points shown in Figure 9.414 to highlight soft skills that are
important across a range of job roles. Regardless of whether some-
222 ARTIFICIAL INTELLIGENCE FOR HR

FIGURE 9.4 Future soft skill importance by role

Business executives

People managers

Talent development leaders

0% 10% 20% 30% 40% 50% 60% 70% 80%

Collaboration skills Communication skills Leadership skills

SOURCE Adapted from CLO Media

one holds a role as a senior executive or a line manager, these soft

skills are continuing to increase in importance over time. The model
brings together these and other disparate sources and research efforts
in academic and industry environments to paint a picture of what
skills will matter most in the workplace of tomorrow.

Creativity
AlphaGo is the name of an algorithm trained to play one of the
world’s oldest board games – Go. Go is a strategy game that might
remind you a bit of chess in that it’s deceptively simple to learn but
challenging to play. The object of the game is to take over sections of
the board, but at the same time your opponent is attacking, counter-
ing and strategizing to outmanoeuvre you. However, unlike chess, Go
has an exponentially greater number of game permutations. That’s
why the algorithm’s creators, Google’s DeepMind AI lab, focused on
Go as their target game. If a computer could win at Go then it should
theoretically be able to play virtually any rules-based game. In a
public exhibition with $1 million in prize money on the line, AlphaGo
challenged Korean world champion Go player Lee Sedol in a five-
game series.
 HR SKILLS OF THE FUTURE 223

The play-by-play of this story is well chronicled online and in the

documentary AlphaGo, but (spoiler alert) the algorithm ultimately
defeated the best Go player in the world in four out of five games.
The only game won by Sedol was due to a move that commentators
could only refer to as a ‘brilliant move’ that would ‘take most oppo-
nents by surprise’.15 When it comes to rules-based activities, whether
it’s a game or a business problem, algorithms have the edge. The
more creativity factors into the equation, the more likely a human
can outperform an AI-based technology.
Being creative is something that humans learn from a young age.
Here’s an experiment you can try yourself: ask any five-year-old you
know to create something, and they may come back to you with an
original song, a painting or even a dance. However, it’s something we
‘unlearn’ as we age. Ask the average 30-year-old to create something,
and they will probably shrug and say they’re not that creative. In
truth, everyone has a creative bent in some form, but we learn to hide
those things (singing in the shower, anyone?) as we age. However, in
a world of increasing automation, creativity is going to be a critical
edge for humans.
Consider this in a business context. When your company is trying
to price a new product, having an algorithm that can examine histor-
ical pricing and predict price points is not a bad way to go. But if
your business is facing new threats from competition, market
constraints or political pressures, an algorithm probably won’t cut it.
Rarely does an issue in the business world have clearly defined param-
eters and a plentiful supply of comparison data. Creativity is the
solution to myriad issues, including this one.
Earlier in this chapter I mentioned Egan Cheung’s take as a technol-
ogy expert on how jobs will change fundamentally over time. However,
one of his comments in our discussion was counter to what I would
expect to hear from someone with an engineering degree and twenty-
plus years of putting that expertise into use as a technology leader.
When I asked him about what will change from a skills perspective, he
quickly replied, ‘Imagination. The future is coming at us so fast, but
maybe imagination is at the heart of it. The ability to dream is the
source of every human advancement. Everything we’ve accomplished
224 ARTIFICIAL INTELLIGENCE FOR HR

comes from someone dreaming big and then rallying groups of people
around them to achieve that.’ The idea that dreaming and imagination
can create value is a very real concept. Need an example of why crea-
tivity matters in a business context? Individuals scoring higher on a
scale of innovativeness lead businesses that see higher profits, revenues
and job creation. A few years ago, researcher Dennis Stauffer set out
to examine the practical impacts of innovative leaders. His research
analysed entrepreneurs for their innovativeness and found that those
reaching higher on the scale led businesses that were considered to be
‘home runs’ by investors.16 Home runs were defined as those making
at least $1 million in annual revenue. Additionally, firms with more
leaders scoring higher on the index saw thirty-four times more profits
and seventy times higher revenues.
In a research study by the University of Phoenix, it was found that
managers overwhelmingly want their employees to be innovative and
creative, yet they perceive a gap between the innovative workforce
they envision and the realities of the teams they have.17 Unsurprisingly,
employees in the study also said they want to be creative, but they
feel like their employer could do more to support the growth of crea-
tivity and the associated skills. Creativity and innovation are critical
and necessary to survival in the business world, and they may be the
recipe for thriving in a changing job market. Research from MIT
actually shows the labour market globally shifting away from middle-
skill roles towards the ends of the spectrum of high and low skills,
both of which require a degree of situational adaptability.18 That situ-
ational adaptability is essentially the ability to navigate complex
situations and resolve problems on the fly, and it can happen when
preparing a legal brief or delivering services as a home health aide. It
comes down to being able to handle non-routine manual tasks as a
key part of the job function.
Want to instil a culture of creativity in your own organization? It
requires an approach that both stimulates and sustains creativity.19
Employers often say they want creativity but then inadvertently take
measures to standardize processes, minimize risk and diminish crea-
tive endeavours. Setting creativity into the core DNA and values of
the firm and its people can help to ensure that innovation doesn’t
 HR SKILLS OF THE FUTURE 225

become just another thing on the ‘to do’ list but that it is an everyday
consideration of employees, managers and executives. From a more
personal perspective, improving creativity can take on a variety of
flavours. Oftentimes I’ve found that interdisciplinary teams and
teamwork can help to create new mental connections and associa-
tions that didn’t previously exist, opening up new ideas and
opportunities for creative thinking. Another aspect is simply learning
how to learn new things. Training isn’t always learning, and learning
doesn’t have to include formal training, especially in a corporate
context. Instead, we should explore topics and concepts that are
interesting to feed our creative desires and help us to continuously
expand our cognitive horizons.
From a more philosophical approach, the convenience of automa-
tion is an ever-present force that can limit our creativity. Convenience
is the reward of increased automation, but struggle is one of the things
that drives innovation. Automating processes to the nth degree could
actually make us worse at innovating. In other words, the light bulb
couldn’t have been invented by continuously refining the candle – it
needed an entirely different perspective. That balance of finding the
right things to automate without taking too much humanity and crea-
tivity out of the process will continue to be a challenge for employers.
Innovation, creativity, imagination – whatever you choose to call
it, there is incredible value and opportunity to leverage this human
trait to remain relevant and competitive both today and into the
future. Is creativity something you learn? Something you are?
Something you do? In reality it’s a little of each of these, and we need
to look for opportunities to improve creative thinking from both a
personal and an organizational perspective.

Curiosity
‘Why?’
If you don’t have a toddler, the question ‘Why?’ might not have as
much significance as it does for the parenting population. That’s
because the word is a way of life – a motto, even – for the preschool-
age child as they explore the world around them. We are instinctively
226 ARTIFICIAL INTELLIGENCE FOR HR

programmed to try to understand the world around us. We are

­inherently curious, and that gives us an edge. Why? Because algo-
rithms do what they’re told. If you run a computer program, it runs
until it concludes with a finding within its parameters. However,
humans are quick to ask a very simple question: ‘Why?’ And that
simple query can drive us to find answers that may not lie on the
surface or uncover questions we didn’t even know we had. Humans
are very well designed for handling things like nuance, poorly defined
problems and ambiguity. If we don’t have the answers, you better
believe we’ll find them.
John Medina, a brain researcher, spells it out plainly in his book
Brain Rules: ‘Humans are powerful and natural explorers.’20 As
Medina explains it, we are all hard-wired to be explorers in a migra-
tory sense, but that need can be just as mental as it is physical. Curiosity
is similar to creativity in that it is built into us as humans from our
time as infants. Babies are always trying things to see what happens,
steadily learning about the world around them over time. As we age,
some of us may experiment less often because we have more ‘knowns’
in our lives: the stove is hot, the street is dangerous, and so on. But
work, especially knowledge work, requires some degree of curiosity to
be successful. From a practical perspective, companies can hire some-
one to follow a process, or they can build an algorithm to follow a
process. Over the long term, the algorithm will be cheaper to operate
and less prone to mistakes. It’s no different from assembly line work,
which has already been disrupted by robotics over the years. However,
the more ambiguity and nuance that is involved, the more likely curi-
osity will help workers to differentiate themselves from a self-contained
algorithm carrying out a predefined task.
Curiosity is more than just asking questions to learn. It’s about
asking questions, learning the concept, and then applying it to deliver
better results. From a skill-building perspective, being more curious
can start with asking more questions. Consider this practical exam-
ple. When I speak to executives about growing their talent analytics
competencies, I encourage them to be curious. Unlike recruiting or
talent management, which have pretty familiar routines and work-
flows, finding the right analytics to solve a problem or explore an
 HR SKILLS OF THE FUTURE 227

issue is not as simple. Where will the data come from? How do we
know it’s connected to the problem we’re trying to solve? Does it
predict what we think it does, or is the result being confounded by
other variables? That’s where curiosity can be incredibly important,
and it simply comes in the form of asking a lot of questions. Why do
we measure this? How long has it been that way? Why was that deci-
sion made? What would happen if we stopped? Asking those types of
open-ended questions of varying audiences will uncover different
opinions, perspectives and ideas that can make us more valuable as
leaders. It’s not about trying to uncover an issue to pounce on – it’s
about trying to understand how the world works, because sometimes
we don’t have all the answers (even if we think we do).
Additionally, curiosity should be seen as a natural option for
extending learning to improve performance. For instance, in recent
years a concept called ‘unschooling’ has arisen, which allows student
interests to drive learning across various topics. The unschooling
trend lets curiosity drive student learning, and it can be equally
applied to adults. A personal example came a few years back when I
realized I understood the general concept of calculating return on
investment for a program but didn’t have a firm grasp on the full
model, calculation and method for gathering data and inputs and
generating a thorough report of outputs and intangible benefits. I
spent a dedicated block of time each week over the course of a month
to research the concept and practise applying it in a variety of exam-
ples. Self-directed learning is an outward extension of curiosity and
goes beyond the ‘just Google it’ mentality that each of us utilizes
today when we don’t know the answer to a question. Grabbing a fact
from the internet isn’t the same as learning an abstract concept and
how to apply it. The ultimate goal of learning within the workplace
context isn’t just to increase knowledge – it’s to increase performance.
It’s fitting, then, that research into the types of curiosity that exist
demonstrated that being curious can improve performance.21 Two
varieties that exist include general and specific curiosity.
A concept that has been studied more rigorously in recent years is
curiosity quotient (CQ). Unlike intelligence quotient (IQ), which is
fairly static and can’t be trained in workers, CQ can be, to some degree.
228 ARTIFICIAL INTELLIGENCE FOR HR

In one study researchers were able to show that intellectual curiosity

actually impacts academic performance because there is some innate
drive to understand concepts that may be unfamiliar. In other words,
there’s an insatiable ‘need to know’ that pushes some individuals to
explore questions, unravel mysteries and solve problems.22 This idea
that curiosity is not just something that diminishes as we age but some-
thing that is innately human and potentially a performance enabler is
incredibly valuable. One leader actually used curiosity as a way to
bring out the best performance in his own workforce.
In the book Turn the Ship Around!, David Marquet tells the story
of taking over the worst ship in the military and turning it into one
of the best performers in just a few short years.23 He explains a wide
variety of leadership concepts that led to the ultimate successful
results, but one of them that sticks out is the idea of being curious
instead of being questioning as a leader. His rationale for this is
simple: as a leader, when you start questioning your staff they can
become nervous or overwhelmed. Imagine that your boss came to
you asking why you did this or how you arrived at that solution – it
could become stressful very quickly! However, when you approach
the situation from a curiosity perspective, you can help to put them
at ease and create a better dialogue. How much different would it feel
if your leader approached you openly to ask about how a process
works or what kinds of lessons learned might be garnered from a
recent project. The feeling of genuine curiosity is a truly different
approach for many leaders, and as Marquet’s example shows, it can
lead to different results as well.
As if these examples weren’t enough, data from Harvard shows
that individuals with greater levels of curiosity are less likely to fall
prey to bias.24 When we are curious, we’re more likely to push back
on those things like stereotypes or broad characterizations that others
just accept to be true. Bottom line: curiosity is an incredibly impor-
tant concept that lends itself to a variety of applications across the
workplace and across our lives. Don’t just accept that curiosity is a
part of life – embrace it and encourage it in others.
 HR SKILLS OF THE FUTURE 229

Compassion
Recently, my wife and I went to the movie theatre to see a film. The
feature was an inspiring one, ultimately leading to a triumphant and
celebratory ending. However, it was interspersed with bouts of
sadness and emotion scattered throughout. I’m not normally one to
get choked up by a movie, but in a few especially poignant scenes I
identified with the plight of the character on the screen. Before I knew
it, I had tears brimming in my eyes. This, in essence, is compassion.
Being able to understand and share the feelings of another person is
a fundamentally human characteristic, whether that person is stand-
ing in front of us or on screen, as in my personal example. While few
job descriptions specifically spelled out a requirement for empathy in
the past, it’s one of those skills an algorithm can’t easily replicate and
therefore will increase in demand.
During a recent presentation at the Globoforce Workhuman
conference, Walmart’s Chief People Officer for Corporate Functions,
Becky Schmitt, talked about the company’s embrace of automation.
Like virtually every retail firm, margins are low. However, for the
world’s biggest retailer, low prices are a permanent way of life, not a
periodic sale or discount. One of the areas of focus for its automation
efforts is around cashiers and the checkout process. Customers now
have more options for self-checkout, allowing them to scan and bag
their own groceries. Want to guess Walmart’s fastest growing retail
operation? Spoiler alert: it’s personal shoppers. Instead of spending
money on staffing those cash registers, Walmart is focusing on hiring
and/or training a bumper crop of associates that can do your grocery
shopping for you. Schmitt explained that this experience has actually
been even more meaningful for associates because they are now able
to interact without having to focus on a highly regimented checkout
procedure. The retailer even trains its personal shoppers on how to
handle a wide range of situations during the drop-off process. For
example, how would you react to a young, frazzled mother with a
screaming toddler in the back seat? This scenario-based training
230 ARTIFICIAL INTELLIGENCE FOR HR

enables the associates not only to get the job done but also to be
compassionate and empathetic to the people they are serving. From a
customer perspective, this is clearly a winning proposition.
When an algorithm diagnoses your illness, as in the example that
opened this chapter, who do you ultimately want to explain that
illness – a lifeless computer or an empathetic human being? The chal-
lenge for employers is in how to train workers to be more
compassionate and/or empathetic. For clarity, empathy is about feel-
ing or sharing someone’s pain while compassion is more about
feelings of warmth, concern or care for another individual. An inter-
esting piece of research shows that how the brain responds to training
in each of these two areas is variable.25 Participants in a study were
given empathy training to determine how well they coped with videos
depicting human suffering, and the effects were increased activity in
areas of the brain that process empathy for pain. This is to be
expected, right? The problem comes when someone continuously
maintains a high level of empathy because they face tough situations
on a regular basis, as this can lead to burnout and other negative
emotions. The study was hoping to uncover ideas for how to mitigate
burnout from incorporating the negative emotions of others. The
next step was to determine whether subsequent compassion training
would reverse the negative effects associated with the video consump-
tion, which it ultimately did. What does this mean for businesses in
practical terms? Empathy is important in many roles today. Consider
the Walmart personal shopper example above or just look to health-
care for any number of examples where empathy would play a critical
role. However, compassion training may offer a way to help workers
cope with empathetic stress and improve resilience over time.
This concept of using compassion or empathy isn’t just a feel-good
story. There are times when hard choices have to be made, as
Microsoft CEO Satya Nadella found out when he made the decision
to terminate Microsoft’s smartphone-focused business unit of more
than 18,000 employees.26 In his book Hit Refresh,27 Nadella talks
about how he is pursuing a more empathetic approach to work and
 HR SKILLS OF THE FUTURE 231

life, and shares some deeply personal stories of struggles that helped
to shape him into the man he is today. However, even in challenging
times, compassion can guide us towards choices that need to be made.
In his case, the decision to remove 18,000 employees was made for
the sake of the 100,000-plus other workers that would be impacted
if he didn’t make that call. If the CEO of one of the largest companies
in the world can place empathy in the top skills he wants to hone, the
rest of us can make it a priority as well.

‘BOT’ VS ‘ASSISTANT’: WHICH IS BETTER?

Some companies are taking a very careful approach to the words they use
to describe these kinds of tools because they don’t want to send the wrong
message to users. However, across all the technology firms I interviewed, I
could not find a consistent basis for the decisions. If I were to ask you
which word you prefer to use, what would you say – bot or assistant?
For example, calling an AI tool an assistant seems to be more personal,
but some firms are trying to avoid the sense of AI taking over the role of a
human assistant. On the other end of the spectrum, some companies call
their tool a bot because the word has a bit of a techie ‘edge’ to it. However,
it can feel a little more impersonal.
At the end of the day, the use cases we’ve explored in this book could
be satisfied by either one. However, I want you to consider it through the
lens of your own team and workforce. If you’re rolling out a tool or
system, what term makes the most sense for how you plan to use it?
Additionally, will it have a human name, or will it go by the name of its
provider? As mentioned in Chapter 4, Credit Suisse uses an assistant with
a human name (Amelia) to provide basic IT support like password resets.
Other firms use Evie, Eva, Olivia and other bots to help with meetings,
and the respective bots are named by the companies that created them.
There’s no definite right or wrong answer to this, but it does warrant
consideration. With a heightened sense of concern around automation and
job displacement, even something as simple as a name can change how
people perceive technology.
232 ARTIFICIAL INTELLIGENCE FOR HR

Collaboration
Consider the following scenario. You lead a team of ten sales profes-
sionals and want to hire one more. Each of the current team members
currently sells 10 units a month, or 100 total units. You might hire a
superstar performer who comes in and sells 15 units, 50 per cent
more than everyone else, raising your total to 115 per month.
Alternatively, you might hire a great team player who sells 12 units a
month but also enables the others on your team to sell 12 units each,
raising the team’s overall production to 132 units.
A superstar may not always improve the performance of the rest of
the group. Sport is littered with examples of ‘great’ players hogging
the spotlight and the game, ultimately leading to lower team perfor-
mance. However, a great team player can dramatically improve the
performance of the rest of the group. That level of collaboration is
key, and it will enable workers to be increasingly valuable and
competitive in the future. Collaboration is one of the skills that can
be aided by AI but cannot be replaced by technology. AI can help to
suggest connections or recommend mentors, but it can’t connect you
in a deep way with someone else. If you’ve ever had the strange real-
ization that you’re more connected today than ever before due to
social networking tools, yet you’re simultaneously more secluded
than ever before from interpersonal relationships, that’s exactly what
I mean by this. AI can support connection and collaboration but can’t
replace it.
In the last HR practitioner role I held, I reported directly to the
CEO. I know that CEOs can vary on a spectrum from amazing to
downright awful, but mine was of high quality. He was a legitimate
war hero, but if you didn’t know his story you would never know
that about him. He was humble but also very capable as a leader.
Despite having no prior experience leading a multimillion-dollar
company, he worked with his co-founder to develop a flourishing
business with teams in multiple countries around the globe. At the
same time, this leader was amazing at building teams and influencing
those around him to perform at their highest levels. Time and time
again our ‘small but mighty’ group would compete and win against
larger, better-resourced and more experienced firms.
 HR SKILLS OF THE FUTURE 233

This was backed up in research by i4cp, the Institute for Corporate

Productivity. In one study by the research firm, high-performing
organizations were more likely to focus on collaboration than low
performers.28 The data shows that those firms with the best market
performance are eight times more likely to have leaders that actively
help others build their networks and more than five times as likely
to say that collaboration is prioritized through rewards and recog-
nition. Employers that make collaboration and cooperation a
priority see better results than those that do not. Logically we would
expect this to be true, but the research shows it’s more than just a
buzzword.
In my own firm’s research into performance management practices
we found similar results. Employers with more collaborative cultures
were more likely to say their approach to managing employee perfor-
mance resulted in positive outcomes. That’s a sharp contrast for
employers with competitive or controlling cultures, who were more
likely to say, paradoxically, that their approach to managing employee
performance actually hampers performance and engagement. When
we see our people as troublemakers that need someone to ‘manage’
them on a regular basis, we’re going to treat people in a way that
prevents them from doing their best work. On the other hand, if we
look at our workers as ingredients in a recipe, we can create amazing
results by finding the right combinations of individuals to solve
specific kinds of problems.
Each of these components, whether cultural or managerial,
factors into how employers can build a better atmosphere support-
ive of collaboration. But physical space is a component that impacts
collaboration and interaction as well. New research into social
network analysis and how organizations collaborate internally
sheds additional light on this subject. In a 2018 conference presen-
tation for the Human Capital Institute, Ben Waber, PhD demonstrated
data showing that individuals can be extremely isolated on an
organizational level simply due to physical location. Even when we
assume that email or chat tools can overcome physical space, the
data shows that employees sitting farther away from others are less
likely to interact on a regular basis. His recommendation was to
234 ARTIFICIAL INTELLIGENCE FOR HR

have workers cycle through seating on a periodic basis so that each

individual has a chance to build stronger networks and connections
to new individuals regularly because physical proximity matters,
even in today’s digitized world. One of my favourite things to do
when I worked as the HR leader for an Inc. 500-ranked fast-grow-
ing technology startup was to pack up my essential gear and move
to another floor in the building once a week. I saw other employees
and had important conversations with them that I might never have
had if I had remained in my office on a different level, which I felt
was a critical part of my performance both as an executive and as
the firm’s HR leader.
The Organisation for Economic Co-Operation and Development
(OECD) has done research around the skills that today’s children
need to succeed in life as they grow and mature.29 The findings are
fundamental, looking at components like literacy and numeracy, but
one of the authors of the study also agreed that components such as
‘developing passionate interests, building strong individual relation-
ships, and participating actively in groups… will continue to be
essential to create a meaningful life, no matter what happens with
robots and computers’. Put simply: collaboration as a concept and a
practice is not going to be extinct any time soon.
Think back over your career about the people you’ve worked with.
Can you picture a handful of those that you learned from, enjoyed
working with and would work with again in a heartbeat, given the
opportunity? Those are the collaborators, and they are essential for
helping us as humans to perform and excel. These influential indi-
viduals motivate their peers, build deeper relationships and inspire
confidence in those around them. i4cp’s report also points out exam-
ples of how this plays out for a variety of employers. For instance,
Patagonia is an outdoor apparel retailer renowned for its progressive
culture. The company’s Vice President of Human Resources, Dean
Carter, says that purpose is an essential component of collaboration.
It’s not enough to just throw a team together with the demand that
they collaborate – they need to understand the company’s mission,
vision and purpose if they are to work effectively as a group. Carter
says that teams across the firm factor that into their efforts, whether
 HR SKILLS OF THE FUTURE 235

solving a tricky business challenge or taking time together to serve in

a volunteer capacity related to the mission of the firm.
I’ll offer one final example to cement this idea that collaboration is
key. Back in the mid-2000s when the United States was going through
a financial downturn, the car makers were all in trouble. One of the
firms, Ford, hired a new CEO in 2006 to help guide the company
through the rocky times. Alan Mulally was a different kind of leader,
and the firm’s board hoped that his approach would be enough to
help them make it back from the path they were taking. The company
was losing nearly $6 billion every quarter and there was no positive
end on anyone’s mind.30
One of the first things Mulally instituted was a weekly business
plan review meeting where senior executives shared metrics and
insights into each of their areas. At the first meeting, Mulally quickly
realized something was amiss. Despite the company’s impending
potential bankruptcy, everyone’s metrics looked green, or positive.
He stopped the meeting and reminded the leaders that it was their job
to turn the business around, and that couldn’t be done with false
metrics that didn’t reflect reality.
The following week, the President of Ford Americas brought a red
metric into the meeting, and the rest of the executives held their
breath. They were ready to watch the bloodbath begin as the new
CEO took this guy apart for exposing an issue that could cost the
company millions of dollars to resolve. Instead, Mulally began clap-
ping to show his appreciation for the honesty and courage it took to
bring up the problem. Then, a critical thing happened. Mulally asked
the rest of the team to pitch in their ideas to help resolve the problem
and get the metrics back in line. Everyone began pitching in to help,
support and encourage the leader with their own advice and insights
from facing similar problems in their respective areas.
By the third week, every single executive on his team brought their
own red metrics, hoping to take advantage of the collaborative
opportunity to work with the rest of the team in a trusting, support-
ive environment. Instead of competing with each other, these peers
were now able to work together to solve problems that were affecting
the entire business. While there were certainly other factors in
236 ARTIFICIAL INTELLIGENCE FOR HR

Mulally’s leadership approach that led to positive results for Ford,

this was a critical one as it reinforced from the top down the notion
of collaboration as a business imperative. The firm was able to avoid
bankruptcy and turn the tide in its favour, not just surviving the
downturn but thriving in spite of it.
We can all learn from this and apply the concept back into our own
workplace. Look for chances to connect people. Look for opportuni-
ties to interact with those outside your normal sphere of influence.
Explore options to change your physical space or location. The quest
to break down organizational siloes and collaborate more effectively
has been a staple of business conversations ever since, well, forever.
But, as the data shows, it’s not just about paying lip service to the idea.
True collaboration is a hallmark of a great organization in terms of
both culture and performance, and each of us can start this off in our
own way within our own teams and in our own professional lives.

Critical thinking
According to one research study cited by Epstein in his book Range,
learning to learn is more important than what you learn, at least
initially.31 In one experiment, researchers watched students who were
randomly sorted into Math I, II and III classes at a collegiate level.
They found that the teachers that taught only materials on the test
and helped students the most in Math I created students who strug-
gled in Math II and Math III classes. Conversely, teachers that forced
students to think critically about problems and arrive at their own
rational solutions created students who performed worse in Math I
but better in Math II and III. Translation: learning to learn is slower
and more painful, but it leads to better long-term results. This is a
fundamental aspect of critical thinking, which involves learning how
to think, act and react based on what you perceive.
For nearly ten years, I have worked with HR professionals who
wanted to become certified. Pursuing an HR certification is one way
to demonstrate competency and affirm a commitment to professional
development. I have created training programmes, taught classes and
offered advice to thousands of individuals seeking help with HR
 HR SKILLS OF THE FUTURE 237

certification. In that time I have come to see that the number one
reason people fail the HR certification exams isn’t because of their
experience, time spent studying or any other common measure we
might expect. In fact, it surprises me to see that the people who study
most are sometimes the ones that ultimately fail. The root issue
shouldn’t actually surprise anyone that has actual HR experience: the
students that fail do so because they can’t apply the theoretical
concepts they’ve learned to the practical questions in the exams. For
example, the study text may give you an analysis of how recruitment
marketing works and where it fits into the bigger talent acquisition
perspective, but the question in the exam may ask you about when
recruitment marketing makes sense for an employer and how the
employer should invest funds for maximum impact. This lack of crit-
ical thinking is what ultimately jeopardizes the performance of the
testing candidate, not a lack of knowledge. Critical thinking, as well,
is a challenging skill to develop. However, it’s one of those competen-
cies that an algorithm can’t easily replicate.
From the perspective of machines and humans, we’ll never catch
up with the amount of sheer knowledge a machine can access. When
IBM’s Watson beat Ken Jennings, arguably the world’s best Jeopardy
player, on the game show years ago, there was an interesting shift in
how I personally thought about knowledge versus application. It’s
the difference in knowing the definition of a theory and being able to
think critically about the theory’s concepts to understand when to
apply it to a situation. The balance of theory versus practical applica-
tion and execution is a very common problem for one population:
college graduates. Someone leaving college believes they know what
they need to know in order to get and succeed in a job, right? Actually,
Gallup’s research shows an alarming gap between what educators
think about how well they prepare for the workforce and what busi-
ness leaders think about how much education prepares people for the
workforce.32 According to the study, 96 per cent of chief academic
officers rate their educational institution as effective at preparing
students for the realities of the working world. That sounds positive
until you attempt to reconcile it with the views of the businesses that
hire those former students: just 10 per cent of business leaders believe
238 ARTIFICIAL INTELLIGENCE FOR HR

that graduates have the skills and competencies their businesses need
to remain competitive.
This is due in part to that gap in critical thinking. There’s not an
employer in the world that would ask its candidates to forego critical
thinking or that would deprioritize that concept in the hiring process,
yet businesses so often substitute educational achievement (degree,
grade point average, and so on) in place of any evidence of critical
thought. Even experience on the job doesn’t necessarily prove some-
one’s critical thinking capabilities. The concept from the creativity
component earlier in this chapter of ‘learning how to learn’ is critical
here. We have been conditioned to expect the answer to a question
and have, in some instances, lost the ability or the initiative to crea-
tively think through the problem and solve it. By helping to build that
skill of analysing a problem, understanding the cause and developing
potential solutions, we can help to teach critical thinking as a hard
concept for business leaders. As mentioned in Chapter 8, one of the
challenges of using AI is a lack of context. Even if there is plentiful
data, machines are designed to move from point A to point B.
Machines do not ask whether point B is even the right destination.
In the last few years I’ve run a wide assortment of workshops that
are essentially HR hackathons. In these sessions, I coach business
leaders on how to frame an issue they are facing, break it down into
its smaller pieces and then solve a small piece of the puzzle. We have
plenty of resources helping us with our five-year plans or our annual
strategies, but there are some problems that need a small, immediate
solution. For example, one firm last year was facing issues with its
candidate experience due to some members of the recruiting team not
following the established procedures and best practices. Within 20
minutes we had come up with a handful of ways to resolve the issue,
from using the ‘best’ recruiters as change agents and champions to
using a leader board to highlight the spectrum of candidate experi-
ence scores by name, putting some positive peer pressure on the
recruiters to improve their results. The way you know you’ve arrived
at a good solution in this exercise is that you nearly smack yourself
 HR SKILLS OF THE FUTURE 239

for not thinking of the relatively simple idea sooner. These kinds of
hackathons can help us to practise the critical thinking and reasoning
skills we need to thrive as business leaders.
MIT research points out that leaders who can create very focused
problem statements, a process of defining an issue or challenge in
detail, are actually more productive and perform better than their
peers with less focus.33 Isn’t that interesting? Those individuals that
can define problems best are more likely to perform better than those
that can’t. When you think about it from a practical standpoint, it
makes sense that defining an issue correctly means you can get right
to work on solving it. But how often do employers treat the symptoms
of a problem for years without actually attacking the root cause?
This discussion about soft skills is wrapped up perfectly by a quote
from one of Deloitte’s senior Canadian leaders, Gordon Sandford: ‘If
you train yourself to be a robot, then you’re out of a job.’34 For now
we’re not fighting, literally or figuratively, against robots for most
knowledge worker jobs. But there may come a time when more of the
work that we’re doing is going to be overcome and automated by the
various tools and technologies that are being developed in laborato-
ries around the world. That said, Sandford’s sentiment is one to keep
in mind: differentiate your skills or face the consequences. We have
control over our own capabilities. Instead of feeling helpless or
powerless against a tide of oncoming technological change, we can
brace ourselves by honing those core human competencies that sepa-
rate us from robots and algorithms. Whatever happens with the
advances in AI technology, you’re in control of your own ­development,
growth and capabilities.

A final question
In all of this discussion about people and the technology that supports
them, I have debated with myself relentlessly on a very important
question: does technology make work more human or less human?
240 ARTIFICIAL INTELLIGENCE FOR HR

To be honest, I think it’s pretty clear that I’m a technological o

­ ptimist.
I believe in the power and value that technology can bring. We have
the power to make work more human, not less, with the advent of AI
technologies and tools. Work can finally be more human than ever.
Ironically, one of the most feared advances in technology, the devel-
opment and adoption of AI, is what it ultimately took to get us to this
precipice. This fundamental human act of working, whether as an
employee, a contractor or some other type of resource, is truly enabled
by AI technologies.
Because of the systems we have explored in the pages of this book,
work can be incredibly more human. The ‘black hole’ of resumes,
where candidates apply and never hear a response? Today those
candidates can connect with helpful chatbots that offer them a
conversation, an opportunity to express their interests and a chance
to ask burning questions. And people actually thank the bots for
those discussions! Learning technologies can help us to teach people
in individualized ways that help them learn best – no more cramming
everyone into a classroom for every single training need. And, finally,
we can manage to the individual using engagement and talent
management tools that give us a full view of someone’s true capabili-
ties and interests.
If nothing else, at least consider the skills component we discussed
in this chapter. By honing your critical soft skills, you can prepare
yourself and your workforce for a more automated future. The
advance of technology is inevitable, but your skills and capabilities
don’t have to become outdated, outmoded or obsolete. Focus on
growing your compassion so you can meet the needs of the human
workforce you interact with. Build up your curiosity by asking ques-
tions, exploring unexplored concepts and connecting with new
people and ideas. Expand your collaboration by meeting with other
leaders around the world using virtual tools and platforms while
simultaneously deepening your connections with those people closest
to you in the workplace. Don’t ever feel like you’re powerless in the
face of technological change, even something as powerful and far-
reaching as AI, because you are ultimately capable of managing your
own destiny.
 HR SKILLS OF THE FUTURE 241

KEY POINTS
●●
As automation occurs, the work that remains will become increasingly
‘human’ in nature. Therefore, developing and refining core human skills
will be necessary to long-term stability.
●●
Five key areas of skills exist that help to set us apart from what
machines can accomplish: creativity, curiosity, compassion, collaboration
and critical thinking.
●●
Above all else, technology enables work to become more human. And,
as business leaders, HR executives and those that care about the
workforce, this is a very positive outcome.

Notes
1 C Wanjek. AI boosts cancer screens to nearly 100 percent accuracy, Live
Science, 21 June 2016. www.livescience.com/55145-ai-boosts-cancer-screen-
accuracy.html (archived at https://perma.cc/3758-W8C9)
2 K Gilchrist. There’s one trait tech companies are after – and it has nothing to
do with IT, says PayPal’s global head of people, CNBC, 15 May 2018. www.
cnbc.com/2018/05/15/technology-careers-top-traits-for-success-from-paypal.
html (archived at https://perma.cc/VX6X-YMVR)
3 M Simon. Watch Boston Dynamics’ SpotMini robot open a door, Wired, 2
December 2018. www.wired.com/story/watch-boston-dynamics-spotmini-
robot-open-a-door/ (archived at https://perma.cc/A28J-Q6QX)
4 Associated Press. Japan’s robot hotel: A dinosaur at reception, a machine for
room service, The Guardian, 16 July 2015. www.theguardian.com/
world/2015/jul/16/japans-robot-hotel-a-dinosaur-at-reception-a-machine-for-
room-service (archived at https://perma.cc/B26K-ZXTR)
5 S Ransbotham, S Khodabandeh, D Kiron, F Candelon, M Chu and B
LaFountain. Are you making the most of your relationship with AI?, BCG, 20
October 2020. www.bcg.com/publications/2020/is-your-company-embracing-
full-potential-of-artificial-intelligence (archived at https://perma.cc/FK96-MHB5)
6 P Caughill. Nearly half of current jobs could be automated by 2055, according
to a new report, World Economic Forum, 3 February 2017. www.weforum.org/
agenda/2017/02/nearly-half-of-jobs-could-be-automated-in-the-future-heres-
what-the-researchers-are-saying (archived at https://perma.cc/Z4U6-ZDDZ)
7 D Epstein (2019) Range: How generalists triumph in a specialized world,
Macmillan, New York
242 ARTIFICIAL INTELLIGENCE FOR HR

8 SD Worx. UK HR professionals fear that artificial intelligence will lead to job

losses in their sector, Press Releases, 12 December 2017. www.sdworx.co.uk/
en/press/2017/uk-hr-professionals-fear-that-artificial-intelligence-will-lead-to-
job-losses-in-their-sector (archived at https://perma.cc/3QG9-34QN)
9 A Smith and J Anderson. AI, robotics, and the future of jobs, Pew Research
Center, 6 August 2014. www.pewinternet.org/2014/08/06/future-of-jobs/
(archived at https://perma.cc/8S33-ECEZ)
10 A Bellos. He ate all the pi: Japanese man memorises π to 111,700 digits, The
Guardian, 13 March 2015. www.theguardian.com/science/alexs-adventures-in-
numberland/2015/mar/13/pi-day-2015-memory-memorisation-world-record-
japanese-akira-haraguchi (archived at https://perma.cc/Z9VC-WCZ4)
11 B Eubanks. We asked 800+ employers about virtual recruiting, Lighthouse
Research & Advisory, 6 April 2021. https://lhra.io/blog/report-we-asked-800-
employers-about-virtual-recruiting-new-research/ (archived at https://perma.cc/
C7BC-FQJU)
12 M Chui, J Manyika and M Miremadi. Where machines could replace
humans – and where they can’t (yet), McKinsey Quarterly, 8 July 2016. www.
mckinsey.com/business-functions/digital-mckinsey/our-insights/where-
machines-could-replace-humans-and-where-they-cant-yet (archived at https://
perma.cc/XK8K-HBH3)
13 J Newfield. Overcoming obstacles to find truth, Thrive Global, 23 July 2021.
https://medium.com/thrive-global/preparing-our-kids-for-an-ai-world-
neuroscientist-on-why-social-skills-matter-now-more-than-ever-e42702d3dc1b
(archived at https://perma.cc/58FC-PLTS)
14 M Prokopeak. The digital future is human, Chief Learning Officer, 7 May
2018. www.clomedia.com/2018/05/07/digital-future-human/ (archived at
https://perma.cc/876X-TEJ2)
15 C Metz. Go Grandmaster Lee Sedol grabs consolation win against Google’s AI,
Wired, 13 March 2016. www.wired.com/2016/03/go-grandmaster-lee-sedol-
grabs-consolation-win-googles-ai/ (archived at https://perma.cc/5T9A-VA4Q)
16 D Stauffer. Personal innovativeness as a predictor of entrepreneurial value
creation, International Journal of Innovation Science, 2016, 8 (1), 4–26. www.
emeraldinsight.com/doi/full/10.1108/IJIS-03-2016-001 (archived at https://
perma.cc/7GD5-X6DH)
17 https://upstarthr.com/university-of-phoenix-research-your-employees-arent-
innovative-enough-podcast/ (archived at https://perma.cc/RW4S-ZL8V)
18 D Artor. The Polarization of Job Opportunities in the US Labor Market, The
Hamilton Project, Centre for American Progress, 2010. https://economics.mit.
edu/files/5554 (archived at https://perma.cc/U5YA-H2FK)
 HR SKILLS OF THE FUTURE 243

19 E Martins and F Terblanche. Building organisational culture that stimulates

creativity and innovation, European Journal of Innovation Management, 2003,
6 (1), 64–74. www.emeraldinsight.com/doi/full/10.1108/14601060310456337
(archived at https://perma.cc/T7JK-25LF)
20 J Medina (2008) Brain Rules: 12 principles for surviving and thriving at work,
home, and school, Pear Press
21 JH Hardy, III, AM Ness and J Mecca. Outside the box: Epistemic curiosity as
a predictor of creative problem solving and creative performance, Personality
and Individual Differences, 2016, 104, 230–37. www.sciencedirect.com/
science/article/abs/pii/S019188691630900X (archived at https://perma.cc/
EZV2-DB24)
22 S von Stumm, B Hell and T Chamorro-Premuzic. The hungry mind:
Intellectual curiosity is the third pillar of academic performance, Perspectives
on Psychological Science, 2015, 6 (6), 574–88. www.hungrymindlab.com/
wp-content/uploads/2015/10/von-Stumm-et-al-2011.pdf (archived at https://
perma.cc/XN2T-48QT)
23 LD Marquet (2013) Turn the Ship Around! A true story of building leaders by
breaking the rules, Portfolio Penguin, London
24 F Gino. The business case for curiosity, Harvard Business Review, September–
October 2018. https://hbr.org/2018/09/the-business-case-for-curiosity
(archived at https://perma.cc/QGK7-4NR5)
25 OM Klimecki, S Leiberg, C Lamm and T Singer. Functional neural plasticity
and associated changes in positive affect after compassion training, Cerebral
Cortex, July 2013, 23 (7), 1552–61. https://pubmed.ncbi.nlm.nih.gov/
22661409/ (archived at https://perma.cc/7UKZ-QWAS)
26 A Shahani. How do you turn around a tech giant? With empathy, Microsoft
CEO says, NPR, 25 September 2017. www.npr.org/sections/
alltechconsidered/2017/09/25/553431516/how-do-you-turn-around-a-tech-
giant-with-empathy-microsoft-ceo-says (archived at https://perma.cc/
HLU7-RJG9)
27 S Nadella, G Shaw and JT Nichols (2018) Hit Refresh: The quest to rediscover
Microsoft’s soul and imagine a better future for everyone, Harper Business,
New York
28 E Samdahl. Top employers are 5.5x more likely to reward collaboration, ic4p
Productivity [blog], 22 June 2017. www.i4cp.com/productivity-blog/top-
employers-are-5-5x-more-likely-to-reward-collaboration (archived at https://
perma.cc/9YPY-ST4F)
29 C Rubin. The global search for education: Are you as good as your robot?
Medium, 11 May 2018. https://medium.com/edmodoblog/the-global-search-
for-education-are-you-as-good-as-your-robot-9a14d117eeb8 (archived at
https://perma.cc/2G8J-WME5)
244 ARTIFICIAL INTELLIGENCE FOR HR

30 E Edersheim. Alan Mulally, Ford, and the 6Cs, Brookings, 28 June 2016.
www.brookings.edu/blog/education-plus-development/2016/06/28/alan-
mulally-ford-and-the-6cs/ (archived at https://perma.cc/X36N-5VQ7)
31 D Epstein (2019) Range: How generalists triumph in a specialized world,
Macmillan, New York
32 B Busteed. Higher education’s work preparation paradox, Gallup, 25 February
2014. http://news.gallup.com/opinion/gallup/173249/higher-education-work-
preparation-paradox.aspx (archived at https://perma.cc/G2H7-BTHA)
33 T Kinni. There’s one question you must ask before solving any problem (it’s
also the most underrated management skill), Inc., 20 June 2017. www.inc.
com/theodore-kinni/theres-one-question-you-must-ask-before-solving-any-
problem-its-also-the-most-un.html (archived at https://perma.cc/9FBH-4RNE)
34 S Harrington, J Moir and J Allinson. The Intelligence Revolution, Deloitte,
n.d. www2.deloitte.com/content/dam/Deloitte/ca/Documents/human-capital/
ca-EN-HC-The-Intelligence-Revolution-FINAL-AODA.pdf (archived at https://
perma.cc/52A3-SPRT)
 245

AFTERWORD

Since I sat down the very first time to start writing this book, the
variety and depth of AI technologies in the marketplace has already
changed dramatically. It’s strange, really, because I feel as if I’m shoot-
ing at a moving target. If I was writing about gravity or some other
universal constant, the research and practice around the topic would
have remained fairly static throughout the writing process. However,
when you’re looking at a new field with cutting-edge developments
occurring virtually every day, it’s a bit different.
The reason I wrote this book was due in part to the need for an
unbiased viewpoint on this concept. So much of the information I see
on a daily basis around this topic is either wildly overhyped, too
theoretical or technical for the average practitioner to understand
and apply, or downright incorrect. As an independent researcher I
have the freedom to explore the topic to my heart’s content and talk
about the good, the bad and the ugly. However, a company develop-
ing these technologies has a vested interest in telling the positive side
of the story, often elevating it to the level of marketing hype.
Additionally, my focus on the practical impacts keeps me from being
too academic in my approach, which is a good thing for the millions
of employees at companies I have supported over the years. They
need more than theory to get the job done, to get the product shipped
and to manage a highly complex workforce – they need actionable
intelligence.
As the range of solutions that leverage machine learning and AI
technologies becomes more prevalent and integrated into the human
capital industry, my hope is that these issues of hype and theory
dwindle away or are at least replaced by relevant and practical infor-
mation that supports those that need it most. Why? Because we need
technologies like this.
246 AFTERWORD

It’s pretty clear where I stand on the matter. Technology has the
power to radically improve the experience for candidates, employees
and the rest of the company if used properly. However, is it smart to
expect that every employer is going to approach this with the right
frame of mind?
As I neared the completion of the book, I saw several news items
criticizing the use of AI because of concerns about bias or a fear of
causing the hiring process to become dehumanized. I have addressed
these items within the context of the book from the perspective of an
HR or business leader, but for candidates and the public at large it
can absolutely seem to be very exclusionary to be shunted off to an
algorithm instead of having the chance to interact with a real person.
It leads me to believe that there is a need for ethical standards to help
guide vendors as they develop the technology and for business lead-
ers as they seek to implement systems that include AI components.
After all, with great power comes great responsibility. I have been
open about my optimism here, and what I feel compelled to point out
is that technology advances have always been and will continue to be
a fact of life. From the first light bulb to the smartphone, technology
is always evolving. The unchanging component is the people behind
the technology.
The truth is that we, as people, have the capability to dehumanize
virtually anything, and that includes processes that have technology
involved or not. People are ultimately the ones that decide how to
treat others, whether they use a piece of technology or do it in person.
At the same time, we have the opportunity to create incredibly posi-
tive experiences for people. Some companies around the world are
known for their ability to create amazing customer experiences that
drive brand loyalty and value for the business. Those firms are able to
create raving fans for their products and services by treating everyone
with respect and appreciation. We need to approach the people side of
the business with the very same perspective. Let’s use the things that
AI can do to scale up and deliver incredibly human experiences that
leave candidates and employees feeling like they received white glove
service, even if it’s being ultimately delivered by a computer.
 AFTERWORD 247

I have come to the realization after years of researching and writing

about technology that it’s usually not about the technology at all.
Funny enough, it really is all about people. I hope you walk away with
the knowledge that in spite of the technological advances and changes,
the revolutionary capabilities, and the continued development and
exploration of AI systems, it still comes back to people. That’s ulti-
mately why I decided to finish the book with a discussion around the
fundamental human skills that we can’t let lapse or diminish in impor-
tance. If we have those critical skills around compassion, creativity
and so on then we can ensure that we’re always on the right path with
our technology decisions. If we only make decisions based on profit
margins, we’ll make choices that take us down an unhealthy path.
One of the recent research projects I had an opportunity to partic-
ipate in required me to envision what AI will look like in the next ten
years. It was an interesting activity that gave me a chance to stretch
my thinking, but ultimately like all of those predictions about jobs
and robots it was no more than an educated guess. What I do know
is that today, there are some great opportunities to leverage technol-
ogy, and some of those technologies have AI within them, extending
their capabilities beyond what we’ve seen historically. Every facet of
our personal lives is being impacted by these ever-evolving systems
and tools, including economically and socially. Work is such a big
part of life for the people that companies employ, and if there’s an
opportunity to use technology to make that work experience more
human to some degree, we should absolutely take it.
Thanks for reading this book and for your commitment to learn-
ing about the opportunities these technologies bring. Above all, stay
human, my friends.
248

THIS PAGE IS INTENTIONALLY LEFT BLANK

 249

INDEX

Italics denote information within figures and tables.

Achievers Listen 173 augmented reality 154–55

administration (core) functions 51–55, augmented writing tools 107, 122
57–58, 61–62, 81–101 automated assistants 61, 76–77, 95–96
see also compliance; payroll automation 17–18, 53–54, 57–63, 164,
adoption rates 191–93, 204–05, 212–14, 217–21
learning 11, 141–42 learning 143
technology 73, 77, 93, 94 recruitment 65, 104–05, 110–12,
advanced machine learning 35–36 114–15, 118–19, 125, 217–21
advertisements, job 107, 112, 122, 124, see also chatbots (bots)
160, 187 automators 212
agility 4, 10–11, 136, 211 Autopilot 42
AI effect 32, 205 average, as measurement 49–50
Alexa 42, 69, 84–85 Axonify 152
algorithm aversion 193–97
algorithm bias 35, 56–57, 185–91 Baby boomers 91
algorithm modifications 196–97 backfire effect 108
Alibaba 42, 44 background screening 44–45, 188, 201
Allie 173 Bailey, Randy 110, 192
AlphaGo 222–23 Beamery 112–13
alumni recruitment 120 behaviour prediction 28–29, 74, 85, 116–17,
Amazon 65, 114–15, 185–86 166–67
see also Alexa benefits administration 90–94
Amelia 95 BenevolentAI 45
analytical skills 58, 210–11 bias 51, 97–98, 107–09, 115, 117
Anchor Trust 126–28 algorithm 35, 56–57, 185–91
anchoring 108 big data 4, 39
Android Messenger 37–38 Boolean search 111
applicant flow analysis 121 boot camps 138
applicant tracking systems 105–06, Boston Consulting Group augmentation
112, 139 model 212
application programming interfaces 201–02 Boston Dynamics 211
artificial intelligence, defined 30 bots see chatbots (bots)
Artificial Intelligence Video Interview bounce rates 129
Act 199 Brighthire 123
assessments, recruitment 7, 107, 115, 117, Brilent 103, 110
118–19, 140 ‘buddy punch’ practice 86
assistants 231 business executives 222
see also Alexa; chatbots (bots); Cortana; business of learning concept 135
Siri
AstuteBot 42 California Consumer Privacy Act 199
asynchronous video interviews 65, 116 candidate (job) matching 71, 84, 99, 107,
attendance systems 75, 76, 82, 85–86 115–17, 123–24, 175–79, 187–88
audits 57, 189–90 candidate receptivity 113–14
250 INDEX

candidate relationship management contractors 160–61

platforms 112–13 core functions see administration (core)
career development (pathing) 153–54, functions
174–81 correlation 27, 165–66, 198
see also internal mobility Cortana 42, 69, 114
Carnegie Mellon 1–2 Covid-19 4
Catalant 176, 180–81 creativity 191, 220, 221, 222–25
causation 27, 28 Credit Suisse 95, 177, 231
Celential 54 critical thinking 192, 215, 236–39
Ceridian 69, 84–85 Cronofy 202
certification 138, 236–37 cultural fingerprint 118
change management 54, 215–16 culture 43, 72, 233
Chapman, Stacey 107 Culture Amp 66–67
chatbots (bots) 72–74, 202, 231 culture fit 118, 140
learning 144–45, 147, 149–50 curiosity 136, 148, 225–28, 240
recruitment 17, 58–59, 60, 104–05, curiosity quotient 227–28
124–30 customer experience 6, 13, 246
talent management 173, 178–79 customization 16, 72, 133–34
workforce management 85, 92–93, 95,
96, 97, 202 data
see also AstuteBot; Tay explicit 34
chatbot conversation topics 129 inferred 34, 35, 64, 111, 153, 190
chatbot handoffs 129 skills 67–68
chatbot peak activity 129 see also big data
chatbot recruitment conversion data breaches 199
rates 128, 129 data collection 45, 219
chatbot session lengths 129 data privacy 56, 104, 198–200
Checkr 201 data processing 219
chess 32, 210, 214, 222 data security 104, 198
China 44 Dayforce Voice 85
Chipotle 177 deciders 212
cloud technology 39, 68, 94–95 decline bias 108
coaching 65–66, 74, 92, 138, 151–52, 169, Deep Dream 42
173, 214, 238 deep learning 38–40, 45
Coleman 76 see also neural networks
collaboration 137, 149–50, 211, 217, 221, DeepMind 222
222, 232–36, 240 Degreed 150
see also Microsoft Teams delta rule 40
college courses 138, 144–45 descriptive analytics 27
communication 28, 68–71, 91–92, 166, 170, design thinking 15
210–11, 222 detail focus 216
see also chatbots (bots) digital exhaust 68
Community Bridges 167 discrimination 83, 97–98, 106, 188, 189
compassion 220, 229–31 disruption 2–5, 55, 57–58, 59, 226
compensation 124, 171–72, 192, 202 diversity 12–13, 64–66, 97–98, 106, 118,
see also pay parity (equity) 119, 122–23, 178, 189–90, 214
competition 1–2 Docebo 143
competitive (controlling) cultures 233 drugs (medication) development 45, 93–94
compliance 8, 52, 82, 83, 87, 117, 142 Dunning-Kreuger effect 107
computer vision 39–40
concurrent validity 119 E W Scripps 96, 202
confirmation bias 107 ecommerce 36, 200
content views 154 economic impact 3, 31, 46
 INDEX 251

EdCast 150 Fortay 118

education sector 119, 144–45, 237–38 Franklin, Benjamin 220
efficiency savings 124
Eightfold 67–68, 190 gamification 107, 119, 122
emotional intelligence (skills) 66, 214 Gen X 91
empathy 229–31 gender parity (equity) 54, 65, 88–90, 115,
employee assistance programmes 92 170, 171
employee check-in systems 75 General Data Privacy Regulation
employee engagement 6, 11, 13, 54, 64, 74, (2018) 199
163, 165–69, 201–02 gifting 72
employee engagement surveys 49–51, gig economy 3–4
167–68 see also Uber
employee experience 10, 12–15, 95, GigSmart 3
141, 215 Glassdoor 38
employee focus (people focus) 14, 15, 56 Gloat 54
employee lifecycle 14, 169 globalization 2, 8
employee recognition 72–73, 162–65, Go 222–23
169, 233 goal setting 6, 9, 169
employee retention 6, 27, 77, 166 Google 34, 36, 69, 114, 141
employee self-service 5, 94–96, 163 Google DeepMind 222
employee sentiment analysis 38, 63–67, 71, Google Jobs 114
167–68 Google Photos 40
Enigma code 23–24 Google Trends 185
Entelo 110, 111 governance 11, 57, 137
entrepreneurs 224 GPT-2 37
Envoy 110, 111 GPT-3 37
Equifax 199 graduate recruitment 65, 114–15, 119,
equity 2, 12–13, 56, 65 237–38
see also gender parity (equity) Guardian Life 59–60
Erba, Claudio 143
ESPN 16 hackathons 238–39
ethics 56–57, 188–89, 246 halo effect 109
evaluators 212 Hamilton, Kristen 107
expectation management 5, 14 Haraguchi, Alex 216
experiential learning 137, 149 harassment 83, 96–98, 155
explicit data 34 hardware technology 26, 71
external hiring 139, 175, 177, 180, 193–94 headwinds/ tailwinds bias 108
external sourcing 109–10 healthcare (medical) sector 31, 81, 84,
91–92, 93–94, 168, 209–10
Facebook 36, 40–41 HealthJoy 91–92
Facebook Messenger 61, 73, 127–28 Henn na Hotel 211
facial recognition 40–41, 44, 45, 64–65, HiBob 201–02
74–76, 86 high performing company practices 6,
feedback, performance 6, 169, 170, 173–74 10–11, 169–70, 172, 233
financial services sector 2, 45, 95, 176–77 see also talent-driven organizations
five whys concept 137 HiredScore 107, 117
Fiverr 3 HireVue 116
FlowXo 127 hiring freezes 175
forced rankings 169 see also recruitment (hiring)
Ford 235–36 HiringSolved 110
formal learning 31, 136, 137, 140–45 home runs 224
see also boot camps; certification; horn effect 109
training HP 154–55
252 INDEX

HR certification 236–37 Johnson, Katherine Coleman 76

HR staff to employee ratios 16 Joonko 97–98
HR strategy 53
hubbot system 125–26 Kaiser Permanente 12
Human AI Decision Support Tool 212, 213 Karp, Athena 107
human evolution 32 KeenCorp 64
keyword extraction 116
IBM 32, 36, 41, 179, 189 Kim, Jim 174
IBM Watson 31, 70, 237 kind learning environments 31, 214
IBM Watson Recruitment 120–21 Koru 107, 118
IHG 140
illuminators 212 labour market friction reduction 3
image identification 39–40 Leadership Actions 151
imagination see creativity leadership skills 222, 228, 232, 235
Immedis 54, 87–88 leadership succession 178
inclusion 12, 64–66, 122–23 leading variables 28
individualization 133–34, 218, 240 LEADx 151
see also personalization learning 10–11, 31, 54, 133–57, 178, 225
industry convergence 2, 3 adoption rates 11
inferred data 34, 35, 64, 111, 153, 190 individualization 240
Infor 76 informal 10
informal learning 10, 31, 136, 137, 148–51 kind environment 214
innovation 6, 103, 211 self-directed 227
see also creativity social 178
Insideboard 54 see also ‘learning how to learn’
instructor-led training 142–45 learning agility 10–11
insurance industry 59–60, 91–92 learning content curation 6, 11, 141–42,
integration services 70, 73, 83, 85, 96, 107, 145–48
114, 122, 173, 200–03 learning content recommendations 153–54
intelligence 31, 32–33 learning content searches 146
intelligence quotient 227 learning experience designers 215
interdisciplinary teams 225 ‘learning how to learn’ 236–37, 238
internal mobility 8–10, 54, 138, 139, learning management systems 141, 143,
176–81, 189 146–47, 153
Internals First 177 LinkedIn 56, 113–14, 120
interview scheduling 202 live video interviews 116
interviews 123 localization 43–44
scheduling of 105, 202 low rules orientation 103–04
structured 107, 109
video 65, 116–17, 119 machine learning 25, 35–36, 63, 66–67,
112–13, 116–17
Jennings, Ken 237 see also Immedis; Joonko; Swoop Talent;
job ads (advertisements) 107, 112, 122, 124, Tay; Tesla Autopilot
160, 187 managers 163–65, 169, 173, 218–20, 222
job application ratios 104 MasterCard 150
job creation 215, 224 Mather, Laura 107
job impact 55, 57–61 media coverage 186–87
job matching (candidate matching) 71, 84, medical sector (healthcare sector) 31, 81, 84,
99, 107, 115–17, 123–24, 175–79, 91–92, 93–94, 168, 209–10
187–88 medications (drugs) development 45, 93–94
job offers 114–15 mentoring 138, 151–52, 178
job samples 109 Messenger (Android) 37–38
job simulations 7, 154 Messenger (Facebook) 61, 73, 127–28
job tenure 8–9, 28–29, 77, 90, 165 #MeToo movement 96–97, 155
 INDEX 253

metrics 165–66, 235–36 performance reviews (discussions) 66, 170,

Microsoft 33, 36, 44, 68, 114, 230–31 171, 174
see also Cortana; Tay performance support 135–36, 152–54
Microsoft Teams 61, 85, 149, 173, 201 personality 140, 221
millennials 91 personalization 8, 16–17, 27, 54–55, 91–93,
mobility 8–10, 54, 138, 139, 176–81, 189 133–34, 147, 151, 217–18
Modern Hire 191 see also individualization
modifiable algorithms 196–97 pessimism 108
Moravec’s paradox 32 Phenom 43–44, 54, 202, 213
mroads 116–17 Pilot Flying J 94–95
Mulally, Alan 235–36 platform gig economy 4
Mya Systems 54 Polli, Frida 107
predictability 213–14, 218–19
Nadella, Satya 114, 230–31 Prediction Machines (Agrawal et al) 31, 46
natural language processing 37–38, 42, predictive analytics 25–29, 109, 114,
97–98, 104–05, 121–24 120, 134
negative reciprocity 161 predictive modelling 28
Netflix 200 prescriptive analytics 27, 166
neural networks 40–41, 42, 44 prioritization 62
Nexus AI 175–76 proactive initiation 69
proactive responses 69
OakNorth 45 problem solving 214
Oleeo 190 problem statements 239
on-demand pay 54, 86–87 productivity gains 61, 76–77, 96
on the job learning 138 project management 175, 176, 179, 180–81
onboarding 8, 71, 74 Proxfinity 71
one-on-one discussions 166 PwC 141
Onfido 45 Pymetrics 107, 118, 188
onQ 150
open enrolment meetings 90–91 racial justice 2, 12–13
open-text responses (reviews) 38, 50–51, RamcoHCM 73
64, 168 recognition 72–73, 162–65, 169, 233
OpenAI 37 recommenders 212
optimism 108 Recruiter 113–14
recruitment (hiring) 29, 58–59, 65, 98,
Pañña 116–17 189–90, 199, 217
Paradox.ai 60 assessments 7, 140
Parkinson’s Law 18, 63 external 139, 175, 177, 180, 193–94
Patagonia 68, 234–35 graduate 237–38
pay parity (equity) 54, 65, 115, 171 see also background screening; facial
Payactiv 54 recognition; job ads (advertisements);
payroll 58, 75, 82–83, 86–90 Mya Systems; Paradox.ai; Phenom;
see also Immedis talent acquisition
payroll administrators 87 reductionism 190
payroll error costs 86, 87 remote work 2, 4–5, 210
PayScale 172 Replicon 85–86
people (employee) focus 14, 15, 56 Restless Bandit 111–12
perceived fairness 190 resume blinding 106, 190
Perception 151, 167 retail sector 75–76, 84, 152, 200–01
performance feedback 6, 169, 170, see also ecommerce; Patagonia; Wal-Mart
173–74 retention rates 6, 27, 77, 166
performance management 6, 54, 82, revenue 6, 224
169–74, 180–81 ridesharing 3, 15
performance rankings 169 see also Uber
254 INDEX

risk management 45, 83 communication; creativity; critical

robot coaches 151–52 thinking; curiosity
robot play dates 59–60 software 26, 71
robotic process automation 58–59 (technology) adoption rates 73, 77,
robotics 1–2, 58–60, 81, 151–52, 211 93, 94
Royal Dutch Shell 176, 179–81 SourceCon Grandmaster Challenge 103,
rules orientation 103–04 110, 192
sourcing tools 109–15
Salary.com 172 specialization 214
Salesforce 65–66, 68, 89 Spoke 96
Sana Labs 54 Spot 97
School of One 133–34, 151 stacked rankings 169
screening, background 44–45, 188, 201 Stanford One Hundred Year Study on
screening capability (chatbots) 105 Artificial Intelligence 30, 46
search engine optimization 36, 56, 110–14, star ratings 154
146, 212 Starmind 68
Sedol, Lee 222–23 stretch assignments 9
self-development 148 StriVR 54
self-directed learning 147, 148–49, 227 structured interviews 107, 109
self-driving cars 1, 39–40, 42, 43, 52, 192 sunk cost bias 108
self-handicapping 107–08 Sunlight 147
self-interest 149 supervised machine learning 35, 36
self-service tools 5, 94–96, 163 surge pricing 90, 162
self-serving bias 107 surveys, employee engagement 49–51,
sentiment analysis 38, 63–67, 71, 167–68 167–68
70-20-10 model 137 switch-offs 162
sexual harassment 155 swivel chair automation 58–59
SharePoint 68 Swoop Talent 107, 120
Shaw, Julia 97 Syndio 54
Shell (Royal Dutch Shell) 176, 179–81
Shell Opportunity Hub 180 Tabei, Junko 97
shift swapping technology 85 tailwinds/ headwinds bias 108
simulations 7, 154 Takeda Pharmaceuticals 12
Siri 42, 69 talent acquisition 7–8, 54, 103–31
situational adaptability 224 see also recruitment (hiring)
SkillFinder 175 talent development leaders 222
Skills Cloud 68 talent-driven organizations 25
skills data 67–68 see also high performing company
skills development 10, 66, 135, 191–93, practices
205, 209–44 talent management 159–83
see also upskilling talent mobility profit chain 9–10
skills gap 9–10, 138–40, 152 talent rediscovery 111–12
Slack 61, 68, 73, 96, 98, 149, 173 Talent Sonar 107
smart speakers 70 Talvista 190
SMS 61, 73, 105, 125 task prioritization 62
Snyder, Kieran 107 Taskrabbit 3
social learning 137, 149, 178 tax calculations 88
social media 59, 125, 127, 162, 187 Tay 42, 203, 204
see also Facebook; Facebook Messenger; teams
LinkedIn development of 179–81
social skills 214, 221 interdisciplinary 225
Socrates.ai 73, 96, 202 size of 166
soft skills 211, 216–39, 240 technology (software) adoption rates 73, 77,
see also agility; collaboration; 93, 94
 INDEX 255

telemedicine 93–94 venture capital investments 42, 54–55

tenure 8–9, 28–29, 77, 90, 165 verbal communication 68–71
Tesla Autopilot 42 Veterans’ Employment and Training
testing (chatbots) 128 Service 67–68
text analysis tools 67 video interviews 65, 116–17, 119
text mining 38 video streaming services 36, 153, 200
Textio 107, 122 views 154
tier zero support 152 virtual reality 54, 154–55
time creep 18 virtual work 2, 4–5, 210
‘time intelligence’ 86 Visier 98–99
time management 213 voice-based assistants 42, 68–70, 90
time tracking systems 82, 85–86 see also Alexa; Cortana
training 10–11, 136–37, 138, 139, 155, 230
Training Orchestra 142–43 Wal-Mart 135, 229–30
translation 38 WalkMe 74
transparency 56, 106, 124, 161, 180 Watson 31, 70, 237
Turing, Alan 23–24 Watson Recruitment 120–21
Turing Test 24 wearable technology 70–71
wicked learning environments 32
Uber 1, 2, 65, 89–90, 160–62 Willis Towers Watson 172
UberPOOL 161–62 Wonolo 3
UberX 161 work location 233–34
UKG 66, 85, 151 see also virtual work
Ultimate Software 85, 167, 204 WorkCompass 174
Unilever 54, 65, 119 Workday 68, 70
United Kingdom 44–45 Workday Skills Insights 68
Unleash 125–26 workforce management 83–86
unpredictability 214, 218–19 workforce planning 98–99
unschooling 227 workforce scheduling 61
unsupervised machine learning 35–36 World Bank Group 174–75
upskilling 135, 147, 217
usability 5, 73 Xander system 168
user preferences 153
YouTube 200
value 57, 154
variables 27–28 Zoom.ai 61, 123
256

THIS PAGE IS INTENTIONALLY LEFT BLANK