Research and implementation of RPA within an

organization

Denis Polkhovskiy

Master’s thesis
Digital Business Opportunities
2022
 Abstract
9th of October 2022

Author
Denis Polkhovskiy

Specialization
Digital Business Opportunities
Thesis title Number of pages +
Research and implementation of RPA within an organization number of appendices
67 + 3

The term robotic process automation refers to software that may be efficiently designed
to perform fundamental functions across applications, like humans. The software robot
can be set up as a workflow with many steps and actions, such as getting forms,
sending a receipt message, making sure the documents are complete, filing the form in
a folder, and updating a spreadsheet with the name of the form, the date it was filed,
etc. The fundamental concept is to identify a simple, repeating activity. Examples
include order processing and invoice management. The majority of robotic process
automation is currently "rule-based," which excludes AI. However, some existing RPA
technologies, such as Power Automate's AI Builder, provide AI as an additional
capability.

The main aim of the thesis project is an implementation of RPA within an organization
and triggering the use of it for automating other routine tasks. By introducing this
practice, we would get several benefits, such as increasing productivity and efficiency
and reducing company costs.

Today, besides the challenge of choosing among many available RPA tools on the
market, picking the proper initial process for automation is complex. During this
research, it was found manual processes at the accounting department that can be
automated in a company and implementation of RPA. It was identified potential
business processes for automation. Then it was converted to high-level business
requirements to detailed development specifications. It was used diagrams for
visualizing the process. Test cases were defined for preparing for the final validation
after the actual robotic process automation implementation. Then, it was researched
current RPA tools on the market and made a comparison of them. Based on the
comparison, it was chosen the best which fits a company. Finally, a manual process
was selected for the implementation based on previous parts and automated.

In addition, this thesis can serve as a useful guide for acquiring the fundamental
knowledge necessary for beginning work with RPA. This work permits a deep dive into
RPA, the identification of the optimal manual process, and a comparison of RPA tools.

Keywords
Robotic process automation, Artificial Intelligence, Power Automate Desktop
Contents

Abbreviations .................................................................................................................... iii

1 Introduction ................................................................................................................... 1
2 Company overview and objectives of the research ........................................................ 2
2.1 Company overview ............................................................................................... 2
2.2 Aims and objectives of the research...................................................................... 2
3 Research methods ........................................................................................................ 4
3.1 Action research ..................................................................................................... 4
3.2 Case study ............................................................................................................ 5
3.3 Constructive research ........................................................................................... 6
3.4 Interviews as a data collection strategy ................................................................. 7
3.5 Collaborative data collection strategies ................................................................. 9
3.6 Decisions based on surveyed research methods ................................................ 11
4 Robotic process automation ........................................................................................ 13
4.1 Overview ............................................................................................................. 13
4.2 General drivers ................................................................................................... 15
4.3 General blockers ................................................................................................. 16
4.4 Maturity and adoption level ................................................................................. 17
4.5 Opportunities and challenges .............................................................................. 19
4.5.1 Opportunities and benefits ....................................................................... 19
4.5.2 Challenges and risks ................................................................................ 22
5 RPA tools .................................................................................................................... 24
5.1 Selected tools ..................................................................................................... 25
5.2 UiPath ................................................................................................................. 26
5.3 Microsoft Power Automate .................................................................................. 27
5.4 Blue Prism .......................................................................................................... 28
5.5 Automation Anywhere ......................................................................................... 28
5.6 Decision Matrix for Robotic Process Automation tools ........................................ 29
5.7 Summary of the chapter ...................................................................................... 30
6 The selected manual process...................................................................................... 32
6.1 Summary of the process ..................................................................................... 32
6.1.1 Benefits and objectives ............................................................................ 32
6.1.2 Implementation notes ............................................................................... 33
6.1.3 Background data ...................................................................................... 35
6.2 Process overview ................................................................................................ 36
6.2.1 Applications used in the process .............................................................. 36
6.2.2 High-level AS-IS process map.................................................................. 36
i
 6.3 Detailed AS-IS process description ..................................................................... 37
6.3.1 AS-IS process map .................................................................................. 37
6.3.2 AS-IS process steps................................................................................. 38
6.3.3 AS-IS data descriptions ............................................................................ 40
6.4 Detailed TO-BE process description ................................................................... 41
6.4.1 TO-BE process map................................................................................. 41
6.4.2 TO-BE process steps ............................................................................... 42
6.4.3 TO-BE data descriptions .......................................................................... 44
6.4.4 Business exception handling .................................................................... 45
6.4.5 Reporting ................................................................................................. 46
6.5 Acceptance testing .............................................................................................. 47
6.5.1 Normal test cases .................................................................................... 47
6.5.2 Known business exception test cases ...................................................... 48
6.6 Summary of the chapter ...................................................................................... 48
7 Automation of the process ........................................................................................... 49
7.1 Introduction of Power Automate Desktop ............................................................ 49
7.2 General actions ................................................................................................... 53
7.3 RPA process overview ........................................................................................ 58
7.4 Summary of the chapter ...................................................................................... 61
8 Discussion................................................................................................................... 63
Sources ........................................................................................................................... 66
Appendices ...................................................................................................................... 68
Appendix 1. Extract of implemented main RPA flow source code................................ 68
Appendix 2. Detailed AS-IS process map .................................................................... 69
Appendix 3. Detailed TO-BE process map .................................................................. 70

ii
Abbreviations

RPA Robotic process automation

AI Artificial Intelligence

PAD Power Automate Desktop

iii
1 Introduction

Innovation is the key to success for businesses seeking to stay one step ahead of their
rivals. Accounting and finance are essential for innovation regardless of the type of
business.

Unfortunately, legacy operations are impeding the digital transition of basic accounting
operations in numerous organizations. Robotic Process Automation can facilitate the
technological change of these operations.

Accounting professionals are essential to the expansion of a business, as they provide the
appropriate financial data and ideas. They are also responsible for adhering to applicable
business regulations. As this strategic direction has expanded, finance and accounting
professionals have struggled to complete their job-related responsibilities within the
existing operational transactions. Financial experts frequently desire to spend less time on
banal accounting tasks and more time on tasks with greater strategic impact.

An ongoing technology megatrend is driving digital transformation initiatives to increase

company efficiency, decrease business threats, and increase company transparency. This
transformation has introduced numerous complications. This consists of legacy systems,
paper-based documents, non - structured files, natural language interactions, and
document fragmentation. Organizations must manage these challenges to improve their
efficiency in finances and accounting (UiPath 2021).

A typical invoice processing workflow for a human follows several steps. In the first step,
the company receives a paper invoice or an invoice by email. Responsible employers
should approve these invoices and then submit them to an accounting specialist. An
accounting specialist carries this invoice by confirming the approvals. He extracts the key
information from the invoice and then enters it into the accounting system. Thought, for
big enterprises that receive hundreds of invoices daily. The relevant data may be exported
to a spreadsheet and imported into software. When the invoice is added to the accounting
software, it can be paid according to the payment terms. These payment runs are
conducted manually by accounting professionals by grouping invoices for payment,
removing duplicates, and sending money to banks. Prior to robotic process automation, IT
systems concentrated on making business processes more efficient and transparent
through business process management techniques. They simplified the flow but at a
prohibitive cost of development, and even so, it is questionable whether they truly
automated the process. It is a point when robotic process automation comes to help
(UiPath 2021).

1
2 Company overview and objectives of the research

This chapter introduces a company for which research is performed. In addition, the
research objectives are presented, and the research questions have been formed based
on the goals.

2.1 Company overview

The research was done in a company that provides software as a service to insurance
companies for managing their business claims, cases, and different processes. Let us call
this company “A”. The company is the leading provider of building claims software
solutions in the Nordic insurance markets. At the same time, it has a tangible presence in
Australian, Austrian, Swiss, and German markets. It provides software solutions for
service and maintenance companies to digitally manage a various building, repair, and
property maintenance projects. It specializes in both the insurance and construction
industries. It supports most of the leading insurance companies in the Nordic markets, the
property damage restoration companies that serve them, and other service providers.
Company’s solutions increase the transparency, cost efficiency and quality for their
customers and are offered on a Software as a Service basis to deliver cost predictability.

2.2 Aims and objectives of the research

This research offers an innovative approach for the working process in company “A”. By
introducing this practice, the company would benefit from increased productivity and
efficiency and reduce company costs.

The general objective of this study is the implementation of RPA within an organization
and triggering its use for automating other routine tasks. It includes several sub-goals. The
first aim of this study is to find manual processes at the accounting department, which can
be automated in the company and implement RPA. Another objective is to identify
potential business processes for automation. Then convert high-level business
requirements to detailed development specifications is the next aim. Making diagrams for
visualizing the process is another objective. The next step towards the overall goal is the
definition of test cases for the final validation after the actual robotic process automation
implementation. The next objective includes researching current RPA tools on the market,
comparing, and choosing the best that fits the company. Then, the final aim is to choose a
manual process for the implementation based on previous succeeded goals and
automation of this process.

2
To sum up, the content of the thesis includes research of processes, selection of one
process, research of RPA tools and implementation of one selected process with chosen
RPA tool, an overview of consequences after deploying the practice to the working
process.

Based on the objectives, the main research question of this study is:
How can RPA be implemented and set as a regular practice within an organization, and
how implementation would benefit the company from a business point of view?

The research question can be divided into three sub-questions:

1. What RPA tools are on the market, and which best fits the company?
2. What processes in the company are suitable for automation, and which process to
choose as the start of RPA?
3. How would RPA affect the work process after implementation at the company “A”?

3
3 Research methods

This chapter introduces several research methods in general. The end of the chapter
presents how the study would be achieved in practice based on introduced researched
approaches.

3.1 Action research

Action research is participatory research in which the aim is to solve together practical
problems and achieve change. Shortly this research can be described as “learning by
doing”. It aims at improving the situation by actual actions taken by a researcher. The
focus is simultaneously on producing researched knowledge and achieving change in
practice. A typical target is changing people’s or an organization’s actions. The target may
be any phenomenon related to human life or finding solutions to practical problems, such
as technical, social, ethical, or professional. In general, the goal of action research is to
solve a practical problem in the organization and create new knowledge and
understanding about the phenomenon. The action research is problem-based and
intensely focused on practice. Typical characteristics of action research are problem-
centeredness, active roles of both the subjects and the researcher as actors in the
change, and collaboration between the subjects and the researcher (Ojasalo, Moilanen &
Ritalahti 2021).

The research proceeds as a circle of planning, observing, and evaluating. In action

research, every stage is conducted to each other systematically and critically. The
research and development process is cyclical, and it repeats distinct stages. First, the
goals are chosen, the development problem is defined, and the objectives of the work are
set. After this, literature and other source material are studied to find out if this kind of
matter has already been studied. After getting familiar with the subject matter, the
development target and goals may be more precise, and the project plan is confirmed.
The actual work starts by studying and experimenting with what opportunities there are in
practice for proceeding toward the goals. Then the obtained material is analyzed, certain
matters are evaluated, goals are modified, schedules are made more precise,
experiments are done, they are evaluated. Planning, action, and evaluating the action
take turns in the research process (Ojasalo, Moilanen & Ritalahti 2021). These action
research stages can be shown as a spiral. Kemmis and McTaggart’s action research
spiral is illustrated in the figure 1.

4
Figure 1. Action research (Ojasalo, Moilanen & Ritalahti 2021)

The action research participants are at every stage responsible for their actions to the
other members of the group and, of course, to themselves. The developer’s role in action
research is different from the other approaches because, in action research, the developer
is always an active member of the group. In addition, it must be remembered that the
participants are equal regardless of what their position is in the organization. Action
research can take much time as the question is always about making a real change
(Ojasalo, Moilanen, & Ritalahti 2021).

3.2 Case study

A case study is an in-depth analysis of a specific situation. The case can be the whole
company, its department, staff, product, customer group, system, or for example, a
process. The case study is characterized by striving for producing in-depth and detailed
knowledge about the studied case. The case study makes it possible to give space for
understanding the diversity of the phenomenon without trying to simplify it too much. It is
more important to learn much about a restricted target than little about a broad group in

5
the case study. The question is not about how general something is but how something is
possible or how something occurs (Ojasalo, Moilanen & Ritalahti 2021).

There are only a few research targets in case studies, often only one. The case can be an
individual, group of people, organization, event, operation, process, or geographic area.
The most essential is that the target is understood as a specific entity, a case.

The starting point in a case study is usually the case to be analyzed or studied. It is often
necessary to get familiar with the subject matter before knowing what can be asked about
it and the actual development task. After that, it is possible to start asking more precise
questions, which help search for necessary background information. Then, it goes the
development of suggestions or models. The figure 2 illustrates the standard case study
process.

Figure 2. Typical case study process

3.3 Constructive research

Constructive research is an approach where the focus is on real-life problem solving and
innovations. In constructive research, the goal is to solve a problem in practice by creating
new construction, i.e., a concrete output, for example, a product, information system,
guidebook or handbook, model, method, or plan. The constructive research methodology
is characterized by the close integration of theoretical and practical aspects. Trying to
implement the developed solution and analyzing its functionality and utility are crucial
components of constructive research. (Ojasalo, Moilanen & Ritalahti 2021).

As in all other research approaches, it is recommendable that a project based on

constructive research follows a standardized process. The typical process of "scientific"
constructive research starts with identifying primary aims and continues with a look into
literature and other project-related material to be able to define the scope further. The next
step is to summarize the theoretical framework and make plans for data collection based
on that. Data are collected in separate phases and with different methods to get as many
insights as possible before the final product or service is offered to customers. During the
final phases, the report is written, and the results are disseminated both internally in the
organization and externally to customers and other stakeholders. The fact that the

6
prepared construct should be commenced and taken into use is left out from the figure
since it is up to the commissioner to order the project what will happen when the research
process is finished (Ojasalo, Moilanen & Ritalahti 2021). The figure 3 represents a
summary of the typical constructive research process.

Figure 3. Constructive research process

3.4 Interviews as a data collection strategy

Interviews are among the most common data collection techniques used in research and
development. The purpose of interviews can be, for example, to clarify matters or gain a
deeper understanding (Ojasalo, Moilanen & Ritalahti 2021).

There are different interview methods. Choosing an interview for the data collection
method requires reflection on what kind of interview is planned, which is affected by what
information is needed to support development work (Ojasalo, Moilanen & Ritalahti 2021).

Structured questionnaire interview is often considered a category of their own and the
other types a different category. In a structured interview, the questions have been
formulated, and they are presented in the order decided before. Only the content of the
answers is left open (Ojasalo, Moilanen & Ritalahti 2021).

In a semi-structured interview, the questions are formulated beforehand, but the

interviewer can change their order during the interview. If some new questions occur to

7
the interviewer during the interview, they can be asked (Ojasalo, Moilanen & Ritalahti
2021).

In an open-ended interview, the interviewer and the interviewee discuss the theme or
problem agreed on. The discussion is open, and both parties participate actively and
equally. The discussion can also be informal (Ojasalo, Moilanen & Ritalahti 2021).

In an in-depth interview, the interviewer and the interviewee have a confidential discussion
on the topic of interest. The interviewee is encouraged to tell his thoughts about the topic
as openly as possible. The main thing is to achieve the goals aimed at. New and even
important themes often come up during the interview (Ojasalo, Moilanen & Ritalahti 2021).

In focus group interviews, the interviewer gets the views of several participants when in-
depth individual responses are not required. Such an interview can give the interviewer an
indication of whether there are shared views or varying opinions in a group. It can be that
the interviewer indicates whether there are shared views or varying opinions in a group.

Usually, interviews should be recorded as it liberates the interview from observing the
interviewee. After the interviews as the data are usually transcribed. When the interviews
are transcribed, whether the text should be written in literary or spoken language must be
decided. The decision depends on how the data is utilized, especially in the analysis and
reporting stages. Many interviews include both structured and open-ended questions, in
which both quantitative and qualitative methods are used to process the data (Ojasalo,
Moilanen & Ritalahti 2021).

An analysis is started by reading the transcribed or otherwise processed data material

several times. After that, it is classified as striving to find connections to the theory
applied. Then the developer reverts to the whole interview, interpretation, and linking the
phenomenon to the applied theories or outlining the theory again (Ojasalo, Moilanen &
Ritalahti 2021).

The saturation point occurs when the researcher no longer receives or sees added
information when new interviews no longer give essential information (Ojasalo, Moilanen
& Ritalahti 2021).

It is recommended that the interview material should be analyzed as soon as possible

after the interviews when the information is still fresh, and the interviewer can remember it
well (Ojasalo, Moilanen & Ritalahti 2021).

8
3.5 Collaborative data collection strategies

There are numerous collaborative data collection strategies. Typically, several people
create together new viewpoints by applying creative methods. The purpose is to generate
as many ideas as possible and record them. All ideas are not good, but there are usually
more clever ideas in a vast group than in smaller groups (Ojasalo, Moilanen & Ritalahti
2021).

Brainstorming is one of the standard creative problem-solving methods in which ideas are
generated in groups. A group of 6-12 persons led by the facilitator aims at ideating original
approaches or solutions to a problem. Osborn's book Applied Imagination fueled the
spread of group brainstorming as a tool for increasing creativity in organizations (Osborn
1953). The traditional procedure consists of rules that encourage high productivity
(Osborn 1953). These rules express all ideas that come to mind, the wilder, the better and
quantity is wanted, so the better. Criticism is ruled out and improve and combine ideas
already presented (Sutton 1996).

The 8x8 method is a method that resembles the mind map but is a more disciplined
ideation method. The problem is written down on an A4 paper, after which eight papers
are placed around this, and a viewpoint is written on each of these. The resulting eight
viewpoints become new starting points from which eight ideas are generated (Ojasalo,
Moilanen & Ritalahti 2021).

The six thinking hats method is a mixture of ideating and analyzing. The main idea is to
examine the problem from different perspectives and inspire the participants to change
their perspectives. The primary limitation of idea is that it defends the ego, which is
accountable for the majority of practical thinking errors (De Bono 2017). The hats allow us
to consider and talk about things we really cannot normally consider or express without
endangering our egos (De Bono 2017). Hats describe different viewpoints in distinct
colors, which the individual members of the group or the whole group can vary. The whole
group can take turns in adopting different perspectives to the problem, or the whole group
may take turns in wearing a hat in a similar color (Ojasalo, Moilanen & Ritalahti 2021).

Main focuses of hats by color:

− White is neutrality and objective facts.
− Red is emotions, hunches, and intuition.
− Black is critical and cautious.
− Yellow is optimism.
− Green is creativity.

9
− Blue is control.

The basic idea of remote thinking models is selecting some remote thinking model that is
then applied to the problem in a ”crazy meaningful” way. The first step is to develop a list
of the characteristics of this thinking model. For example, if the purpose is to design a new
school rucksack, the remote thinking model can be a comb. Characteristics of a comb are,
for example, teeth, plastic, small size, lightweight, hardness, and glossiness. Each
characteristic is examined, and based on that, a suggestion is made for a new rucksack.
For example, the teeth of a comb can generate the idea of a” toothy” rucksack, which may
generate the idea that the rucksack has plenty of pockets. When all the characteristics
have been examined, there are numerous suggestions for a new rucksack. The final
rucksack is most likely to combine several new characteristics. A remote thinking model
can be selected from everyday life (Ojasalo, Moilanen & Ritalahti 2021).

The Osborn checklist includes questions that help in generating innovative ideas
concerning, for example, an existing product or service. It may make it possible to develop
a new product or service. It helps to get rid of the familiar thinking routes. The goal is to
generate as many new alternatives as possible when the problem and the question are
reflected on. The purpose is not to answer the questions but to open new perspectives for
thinking (Ojasalo, Moilanen & Ritalahti 2021).

Bodystorming is a generic name for many kinds of methods. The connecting factor is the
use of physical activity when the purpose is to understand the users and the context,
which is the development target (Ojasalo, Moilanen & Ritalahti 2021).

Central characteristics of bodystorming:

− Based on ideas, physical prototypes are made immediately, and testing starts in the field.
− The user environment is created in the place of idea-generating, for example by using
maps and staging and acting different events and situations in which the ideated
products are used.
− Based on information collected in user research, users are described as typical user
characters whose roles designer group members take in different situations. It is
assumed that the actor can act in the same way as a real user also in situations in which
observations have not been made, i.e., to use the ideated equipment or service in the
same way as the user.

In using excursion techniques, the focus is on getting distance from stagnant idea
generation. These techniques include, for example, association games, role play, drawing,
and serial stories. Heuristic idea generation techniques are often called heuristics. It is

10
used to systematically generate and evaluate ideas generated for solving problems or
creating development alternatives in different situations (Ojasalo, Moilanen & Ritalahti
2021).

Wish lists suit well for developing products and services as they enable the company to
clarify customers’ expectations and needs. It is not often easy for customers to talk about
their expectations if they are unaware of them. Customers can more easily tell what they
do not get than describe what they would like to have or need in certain situations. The
purpose of the wish list is to find out what customers would like to have if they did not
have any restrictions. It draws attention only to real or imagined benefits. When this
technique is used in group discussions, all comments and ideas are written down on the
board, and no one can criticize them (Ojasalo, Moilanen & Ritalahti 2021).

Crowdsourcing describes a dispersed problem-solving and production model in which the

commissioner utilizes the community’s competence for a specific limited task.
Crowdsourcing refers to a practice in which the company or organization outsources some
activities that their staff had earlier performed (Howe 2006). Activities can be outsourced
to an undefined network of people in the form of a free invitation. This can mean
measures taken by individuals, pairs, groups, or crowds. The critical characteristics of
crowdsourcing are open invitations and an extensive network of participants (Ojasalo,
Moilanen & Ritalahti 2021).

3.6 Decisions based on surveyed research methods

The best research approach for this idea is constructive research because we focus on
real-life problem solving and bringing innovations. As a starting point, research can be
split into two problems, and for both problems, we would use constructive research
techniques.

The first problem is which RPA tool should be selected. The selection would be crucial for
the company and future RPA maintainers and developers because the proper tool
selection is a foundation of future successful practice. The research on this problem
should be done thoroughly. For solving the problem, data about tools will be collected
from various sources to get as many insights as possible before making any decisions.
After it, the retrieved data would be analyzed using a decision matrix with criteria for
showing relation and comparison for analyzed tools.

The next goal is to solve the problem of manual processes by creating concrete output. In
this case, it is a process definition document (guidebook or plan for implementation).
Before collecting the data, we should first identify the problem. In this topic idea, the

11
problem is manual processes. Based on it, we should focus on data collection processes.
Based on the covered information in this chapter, the best one to choose or first to be
used interview. Since the company is small and by defining the right group of people to
interview, we can choose several interview methods or use a mixed interviewing strategy
to gain the best output. First, the process can start with semi-structured interviews. I
cannot entirely agree that interviewees should know questions beforehand, but, in this
case, it would be good to let people think before since results are made based on well-
thought answers. Then perform analyses of interviews. After the first loop of the interview
process, we can perform open-ended interviews to discuss thoughts or ideas that
emerged from the first round and then move to the next step. Finally, in the last round, we
can have a focus group interview with a group of people to discuss the results of previous
steps and indicate if views for results are the same or vary. Otherwise, this step can use a
collaborative data collection strategy known as brainstorming with this group of people,
and I can be a facilitator of this session. After analyzing the previous step, saturation point
can be achieved where new interviews or data would not give new essential information.

The research would be started with a broad scope, but by answering research questions
and progressing, it would be narrowed to one manual process and one RPA software to
implement the practice.

12
4 Robotic process automation

This chapter discusses the central concept of the research "RPA" in detail. It includes an
RPA overview, general drivers, and blockers. Also, we would go through the maturity and
adoption level of RPA and see its opportunities and challenges.

4.1 Overview

The Institute for Robotic Process Automation provides the following definition of RPA. It
defines robotic process automation as the application of technology that enables
employees to configure computer software or a 'robot' to capture and analyses current
software systems for executing a transaction, managing data, initiating responses, and
interfacing with digital software (UiPath 2021).

Previously when people mentioned automation, it was mostly referred to as hard-coded

automation workflows or Application Programming Interfaces calls into software programs.
In comparison with that, RPA imitates a human's steps on top of existing tools. RPA acts
similarly to an individual who reads and assesses information from a document and
transfers it to multiple software systems on their desktop. RPA is a computer-coded
application that can substitute people conducting monotonous rules-based activities, and
it has macros that span both functions and applications (Deloitte 2019).

RPA can do following actions (Deloitte 2019):

− Accessing email and attached files.
− Logging into web/enterprise applications.
− Moving files and directories.
− Copying and pasting.
− Form fillings.
− Ability to read and write to databases.
− Extraction of data from the web.
− Integration with system API.
− Making calculations.
− Data extraction from documents.
− Gathering social media data.
− Following ‘if/then’ decisions/rules.

Robots are able to easily transmit data across tools, emulating operations such as
clicking, typing, and navigating between windows. In addition, these robots may use

13
native and Artificial Intelligence and Machine Learning extensions to enhance overall
performance and learn from past experiences.

With RPA, accounting specialists can automate multiple processes they were earlier
unable to. In order to select the appropriate operations for automation, they should have
some of the following characteristics (UiPath 2021):
− The processes and their underlying applications must be stable
− Rule-based process
− Low exception rate (low variation between processes)
− Highly repetitive, manual, and high-volume operation
− The inputs should be electronic or machine-readable

The future role of managers in accounting departments will shift from controlling people to
generating higher-value tasks, including exceptions and judgments involving human
expertise, from their personnel. The orchestration engine simultaneously monitors
operational performance. Finance and accounting professionals will shift away from data
collection, data entry, and bookkeeping in favor of advising positions, where their
judgement and consulting abilities will enable them to become company partners (UiPath
2021).

Automating simple processes such as invoice data entry and cash applications is an
excellent starting point for a business with RPA (UiPath 2021). Before embarking on the
RPA journey, these activities are ideal for conducting proof-of-concept research.

The most profitable way for automating general accounting processes is to take a more
agile and iterative approach for effective automation of the process (UiPath 2021). RPA,
AI, machine learning, and other advanced cognitive technologies are complementary.
Although RPA alone can perform rule-based, repetitive tasks, incorporating AI, ML, and
cognitive technologies into RPA can improve the process. Cognitive and machine learning
technologies can perform many different actions. It can include obtaining payment
information from invoice photos, extracting data from emails, documents, and bank
transaction messages, and so on. In these situations, RPA is utilized to manage open
receivables using accounting system data, get bank statements and input them into the
accounting system, and automate matching payments to open invoices and closing them.

Using native features, numerous current ERP systems, such as SAP and Oracle, have
significantly automated cash-process handling. They have achieved a level of
automation up to 50 per cent. There are still many exceptions or fallouts that require

14
human effort. With the help of RPA combined with AI and ML percentage of ERP
automation can be increased up to 90 percent (UiPath 2021).

RPA is a disruptive technology that can considerably change how existing industries
function. It replaces the well-established finance and accounting practices because it has
significant advantages over the old way. As the RPA continues to evolve, we can be sure
that it will play a significant role in disrupting many widely accepted processes as
standards today. However, the reality is that the potential for automation extends far
beyond industrial application.

4.2 General drivers

The RPA installations that enhanced the sharp rise of software robots in recent years
have produced a wide range of advantages. According to the "Robotic Process
Automation Implementation Drivers" article (Kedziora, Leivonen, Piotrowicz & Öörni
2021), the investment return varies from 30 to 200 per cent in the initial year. As a result
of the personnel having more significant opportunity to concentrate on the client
relationship, thus customer service is getting better. In addition, staff can profit from
decreased duties, enhancing the value of company interactions with clients. Businesses
can measure the success of the implementation of RPA in terms of newly released hours
for distributing for other tasks. It gives employees relief from monotonous activities. They
can now focus on value-added duties, forcing the organization to plan sophisticated
HR policies to retrain workers and use their free time for self-improvement. The benefits of
RPA fall into four major categories: cost reduction, improved productivity, customer
service enhancement, and employee engagement.

RPA can play a critical role in changing operations for accountants and finance
professionals. Previously, companies outsourced numerous transaction-heavy accounting
tasks to locations with cheaper labor. RPA aligns the benefits of using less paid labor
markets from traditional outsourcing or offshoring key business processes with using
resources on a place. It increases productivity and effectiveness without raising risks or
negatively impacting client service (UiPath 2021).

RPA can supplement and increase employees' work capabilities. It lets people accomplish
more with fewer resources in less time. The robots used in RPA are considerably cost-
effective at handling large repetitive tasks. They can operate all the time without any
exhaustion (UiPath 2021). Even though there are initial practice establishment expenses
and ongoing maintenance expenses, the return on investment with RPA is much higher.

15
Nonetheless, investment return may not be the only driving factor for many accounting
departments. Additionally, RPA offers enhanced risk management and compliance,
greater precision, reduced cycle times, and increased throughput. It is crucial for end-of-
month reporting and accounts receivable, reducing the average time it takes for a
business to receive money for sale. Robots release costly to finance and administrative
resources to focus on higher-value tasks and outputs, enhancing employee satisfaction,
profitability, and consumer experience (UiPath 2021).

4.3 General blockers

The article by Kedziora and others (2021) shows five categories related to risks in RPA
implementation. These categories are executive, technical, change management,
operational, and functional risks.

Numerous business exceptions may be triggered by the robot's speed and the
lousy quality of input information. The methods through which robots perceive GUI
components are already strong. Particularly when engaging with the virtualized
environment, they are insensitive to system difficulties and changes. Robots lack human
rational thinking, and their outcomes are predetermined. Any input or process logic
defects will result in the bot repeating mistakes until someone corrects them. Another
factor limiting the adaptation of RPA is unstructured data. RPA requires structured data,
but in many companies, documents are primarily unstructured, even though some tools
can help structure the data before the start of the use of RPA (Kedziora et al. 2021).

In some cases, handwritten documents can be a common blocker in the adaption of RPA
since it can be a big challenge when there are many of them. RPA should have proper
recognition functionality for managing such documents (Pritzker 2020).

One of the crucial blockers is variation. In case if RPA can manage ninety percent of
transactions in ten steps, but ten percent of this process requires an extra two steps for
completion after step three, then automation should have an exception rule for managing
these steps. Exception handling needs to be added into the workflow to address the
variation in this scenario. Even though these operations may also be automated,
integrating workflows into the robotic process can be time-consuming; therefore, the firm
must determine if investing the additional time is worthwhile (UiPath 2021). It is a point
where the role of humans in the RPA process is fundamental. In such digital
transformation, accounting professionals become experts in exception handlers, whereas
robots handle repetitive and routine tasks. Humans will be required to execute
complicated, judgment-based tasks that RPA cannot perform (UiPath 2021).

16
4.4 Maturity and adoption level

Based on the article “The Five Dimensions of Digital Technology Assessment with the
Focus on Robotic Process Automation (RPA)” (Harmoko 2021), the RPA evolution can be
split into three phases.

In the early phase of RPA (RPA 1.0), The Assisted Robot automates various jobs or
processes on a personal workstation to reduce manual handling time and costs and
improve precision. Unfortunately, the assisted RPA still involves human assistance during
actual RPA process execution.

Robots are installed on multiple desktops in the second phase of RPA (RPA 2.0). Robots
run with less human intervention and autonomously connect into a designated system,
initiate an activity, monitor its status, and finish it. These operations are managed on the
interface, which features many panels for assigning jobs, modifying destinations and
queues, and responding to robot output.

Cognitive RPA is the final stage (RPA 3.0). Innovative end-to-end processes utilize
machine learning, pattern recognition, computer vision, and processing of natural
language to manage lengthy and complex operations.

Figure 4. RPA evolution (Harmoko 2021)

Gartner Hype Cycle is a concise summary of the novel technology, innovation's market
expectation and perceived value. Gartner Incorporation introduced this overview in 1995.
Every year this overview is updated development technology and changing market
expectations. Around ninety Gartner hype cycles describe various technologies, such as
AI, digital marketing, advertising, et cetera.

In Figure 5, it can be seen a modification of the Gartner Hype Cycle with five distinct
zones ranging from one to five. The color is determined by the duration of the entry period

17
and the likelihood of failure. The deeper the color, the lengthier the entry time and the
greater the danger associated with the examined technology. At the beginning of the
cycle, industry reactions to recent technologies ranged from mild interest to euphoria. An
increasing market percentage usually observes euphoria. It occurred when RPA was
introduced in the market until 2018. RPA expanded then, and its location was in the "Peak
of Exaggerated Expectations." People view RPA as a solution for workplace automation
that provides total flexibility and is user-friendly. Nevertheless, this is incorrect, and the
public was dissatisfied since RPA is not entirely automated. In addition, it cannot
recognize actual papers, including hard copies and handwriting. This disillusion has
ushered in 2019's "Trough of Disillusionment." Gartner predicted that RPA would ascend
the "Slope of Enlightenment" by 2020 because of the great hopes for cognitive RPA.
RPA's cognitive capabilities include artificial intelligence, machine learning, and natural
language processing, enabling it to analyze paper files and make effective activities.

It is anticipated that cognitive capabilities can elevate RPA to the "Productivity Plateau."
UiPath and Automation Anywhere have implemented the cognitive RPA. AI Fabric, an
UiPath AI capability, trains a robot to comprehend and read many types of physical
papers in various languages. It enables bots to communicate with screens and engage in
conversations with humans. Additionally, the robot is trained to improve organizational
performance by utilizing process mining and task mining functionalities. UiPath and
Automation Anywhere prepare RPA for a completely automated end-to-end process with
AI capability.

Figure 5. The RPA Position in Gartner Hype Cycle (Harmoko 2021)

18
Based on Harmoko (2021), the best time for implementing RPA is when technology is in
the "Climbing the Slope" phase. It is a prudent alternative strongly advisable due to the
reduced danger and quicker entry period. If a firm has essential skills and a substantial
office automation burden, such as an online shop, bank, or insurance company,
implementing RPA as early as feasible or while the innovation is in its infancy will boost its
market position. However, it is extremely dangerous and time-consuming.

Nowadays, the RPA is now at the "Climbing the Slope" phase. Nevertheless, AI skills in
RPA still need to be improved for reaching the "Slope of Enlightenment" and "Plateau of
Productivity" phases in the Gartner Hype Cycle.

The automation era is here, and RPA can help companies succeed and thrive in this time
of technological innovation and disruption. As shown in this chapter, the adaptation and
maturity of RPA are good enough to consider implementing the technology. The shown
interest in RPA has revealed the tendency to grow significantly worldwide, and the
community of users has grown. Based on the Gartner hype cycle, we can see that the
chasm was crossed over, and now is the ideal time for all businesses to examine RPA as
a company task automation strategy.

4.5 Opportunities and challenges

Robotic process automation is frequently viewed as a panacea for enhancing business

efficiency and lowering expenses. In this section, let us examine the advantages and
disadvantages of RPA.

4.5.1 Opportunities and benefits

In most cases, many companies' accounting and finance departments are usually among
the first to realize the benefits of automation in their regular operations as can be seen on
figure 6 (Deloitte, Financial reporting RPA risks and controls 2021). Furthermore, since
automation eventually affects the entire company, automation of accounting and finance
operations may yield much more extensive economic outcomes (ElectroNeek 2021).

Consequently, RPA might be advantageous for accountants and financial specialists who
oversee activities with a high degree of responsibility and precision. Any of their faults
might result in costly issues for the business.

According to a study by Gartner (2019), a human prone error in the finance process in an
annual average of 25,000 hours of unnecessary rework at the expense of $878,000. RPA

19
can manage irregularities in employee effectiveness, establishing a structure that elevates
specialists to more crucial roles and improves corporate outcomes.

Figure 6. Types of RPA deployment

According to Ernst & Young (2021), the tremendous potential application of RPA is for the
following fields:
− Bank reconciliation process.
− Sales ordering and invoicing.
− Fixed asset management.
− Financial and external reporting.
− Inventory management.
− Receivables management.
− Payables management.
− Financial statement closing.
− Tax planning and accounting.

The most satisfying aspect of robotic process automation is the quick investment return
that occurs once RPA is deployed. Robots do not physically integrate into a company's
information structure, which is considered low-hanging fruit due to its comparatively
simple implementation (ElectroNeek 2021).

Consider the three most prevalent RPA application cases in finance and accounting:
accounts receivable, payroll, and financial reporting.

Accounts receivable is an ideal starting point for automating financial processes. Accounts
receivable uses fewer external data in comparison with accounts payable. Days sales
outstanding is the significant RPA parameter that may be used here (ElectroNeek 2021).

20
It relies significantly on both the payer and the receiver's human factors. If an accountant
forgets to deliver an invoice, the result is more than a monetary shortfall. If it occurs
frequently, it threatens the order-to-cash process and impacts stability. When RPA is
implemented, robots will function like a virtual team. They allow rapid resolving of email
bills. It also goes a consistent cash flow without deficiencies by automating this task.
Besides reducing cash gaps, RPA can assist input information. Also, it helps accountants
to have less juggling in multiple information systems.

Moreover, the following lists illustrate accounts receivable jobs and procedures that can
be automated intelligently (ElectroNeek 2021):
− Customer data setup and management.
− Extracting customer information from various sources.
− Sales quotation and entry generation.
− Invoice generation and distribution.
− Cash application.
− Customer credit monitoring.
− Dispute resolution.
− Follow-ups, reminders, and dunning.
− Credit risk management.
− Chargeback.

In payroll use cases, bots assist in reducing payment delays and mistakes by completing
information input, timesheet checking, and reducing computations.

RPA can assist in several situations involving payroll processing:

− Employee data extraction.
− Data verification across information systems (sick days, business trips, timesheets).
− Generating and approving timesheets.

A secure method to observe financial performance is following profit and loss daily.
Nevertheless, manually editing profit and loss statement reports is tiresome and takes a
lengthy process. Such tasks can be managed by RPA, which can also generate flawless
reports in real-time. The automation makes business processes transparent and
guarantees financial forecasting exactness.

There are several reporting processes where RPA can be helpful to be applied
(ElectroNeek 2021):
− Trial balance and balance sheets.
− Income statements.

21
− Profit and loss.
− Variance analysis.
− Financial close processes.
− Regulatory or management reports.

4.5.2 Challenges and risks

As shown in a previous chapter, RPA offers many benefits to a business. Everything has
two aspects, and RPA has its problems and threats. Typically, RPA is a quick solution. It
can offer instantaneous usefulness but demands complicated omnichannel platforms and
frequent workflow assessment and optimization eventually.

Few RPA technologies presently on the market provide modest features of machine
learning. There is still a long way to go until automation is really intelligent. With AI and ML
incorporated into RPA systems, the capability for automation will grow tremendously.
Unfortunately, we have not yet arrived there.

The next challenge is ownership. It means that the owner of the RPA process should be
defined. Unknowing RPA roles and responsibilities in an automated environment can
create disconnected dots. It causes a lack of ownership and responsibility among different
teams.

Selecting wrong use cases for automation is a typical mistake that challenges RPA
implementation and reduces ROI. It is essential to make a case for a proof of concept
before going forward with automation. The most appropriate processes should always be
selected to begin a successful automation journey. Not all processes are suitable for
automation. Based on standardized and predictive rules, processes should be identified
with clear processing instructions. Processes that demand a high grade of manual input,
structured and repetitive input, involve more human mistakes. These processes are
always good to be started for automation.

Lack of skilled resources can be the next challenge. With the growing popularity of RPA,
the need for skilled labor has been increased. RPA implementation can be blocked if a
team lacks skilled resources.

Another challenge is automating processes end to end. Sometimes RPA cannot wholly
automate some processes, and it requires the use of machine learning algorithms, which
can add extra costs to the company and the project. The best way to manage it is to use
the divide and conquer technique. Such scene action should be split into simpler parts

22
and then automated. Another thing to consider is that the extra costs involved in delivering
end-to-end automation raise cost-efficiency questions.

Lack of support from businesses is another risk. Most RPA implementations fail because
of a shortage of required support from the businesses, like workflow diagrams or other "if"
scenarios or how the bot should process different business rules for different data types.
Businesses should provide a comprehensive list of technical exceptions that the
operations team may encounter during manual processing (CiGen, 2021).

The next challenge is cultural change. RPA implementation needs a cultural and mindset
change starting from the senior leadership. Employees' resistance should be managed
because it is the main reason for the lack of willingness to accept RPA. Misunderstood
information about RPA can often create suspicion among people about losing their jobs.
Before starting the automation project, a company should educate employees about what
software robots can and cannot do. It should be clarified that the bots are to be seen as
helping, not replacing their current work roles. Also, a company should train employees
regularly since the automation era will require them to develop new skills.

Unfollowing RPA best practices can create other risks. If the team does not follow best
practices in RPA implementation, it can result in more invested time debugging the code
and challenging possible reuse of the workflow in the future for teams.

The next challenge is unclear expectations. They always should be reasonable and not be
over-enthusiastic. When a goal is not clear, the team, management, and other
stakeholders involved in the RPA implementation would not be able to measure the
success of the technology, and unclear can frustrate its progress. A practical method is, to
begin with, a clear ranking of business goals and then finding out how RPA can help
achieve them.

Technical ambiguity makes another risk. In some cases, RPA deployment does not bring
expected results because of ambiguity among the technical team. The deployment of
automation can fail when technical questions are not clarified about operating
requirements during the implementation.

Another big challenge is a lack of infrastructure. Satisfactory results cannot be achieved

without proper infrastructure and investments in RPA.

23
5 RPA tools

There are many tools available in the market to implement RPA. Understanding the
essential functionalities of an RPA tool is vital to choose a suitable tool for the company.

Core functionalities of RPA tools include:

− Gather inputs from other systems and decide their course of action.
− Program other bots or software robots through an interface.
− Interact with other systems via API integrations and screen scraping.

It is crucial to select the most suitable tool for business. Before selecting, we should define
our requirements. We can highlight several criteria that seem most valuable for us at that
moment.

Technical features are one of the most important aspects to consider when selecting an
RPA tool. The tool must be platform-independent and should support any type of
application. It should have characteristics like as screen scraping, scalability, and
cognitive capacities. We should check a tool's security features because the software
robots manage confidential data. The extent of security should be ensured before
choosing an RPA tool because essential security measures are always necessary to
avoid the misuse of confidential data or privacy issues.

Before selecting any RPA technology, the total cost of ownership is an additional vital
factor to consider. It relies on elements such as license fees, implementation costs, and
maintenance expenses, among others. It should be ensured that the chosen tool meets
the dedicated budget for the deployment of this practice. In our case, we decided to start
small and then expand to a larger scale when necessary. It would be easier to realize and
filter what tools functionality is needed for the company's business process when working
with the technology. By doing small steps, the company can also minimize expenses on
unnecessary RPA tools and help to invest effectively.

Ease of use is the cornerstone for automation. The chosen tool should be user-friendly
and versatile enough to support fundamental automation procedures. Whichever tool we
choose, we must consider that it does not need much training and is entirely user-friendly.
RPA tools are created to simplify complex operations and automate them to decrease
human mistakes. It should be mirrored in the design of its interface. RPA tools should be
comfortable for all employees, regardless of how comfortable they are with programming
techniques since it will lead to higher levels of adoption. A complex user interface will slow
down implementations and cause novel issues. Good indicators of ease of use are

24
characteristics like image recognition and drag and drop functionality. Hence, the more
accessible the tool to use, the faster the benefit to the company.

Exception handling is the following criteria to consider. The preferred RPA tool must
include a thoughtful procedure for addressing exceptions. Automated error detection
should be more efficient than manual error detection. When errors require human
attention, they should be redirected towards experts. Nevertheless, if those errors can be
managed automatically and there is no need for human experts, it will increase the speed
of all operations.

RPA tools providers should offer strong vendor support to their clients and preferably have
extensive experience. It is always good to choose dependable providers with a good
support system. They should have automation experts and offer specialist advice to
ensure the business does the best with the tool. It is easy to resolve any issues arising
during the implementation and maintenance phases with such support. Extensive support
helps make smoother deployments, increasing technological innovations, training,
certifications, and more.

Reliability is an essential factor while considering an RPA tool. A product with a notable
history on the market and a solid reputation would be a more dependable option than one
recently launched. Communities associated with this technology are another vital aspect.
When a community has a large number of developers and users, this may signal an
abundance of knowledge and answers to issues, and it would be easy to discover
solutions to problems without paying for substantial vendor assistance.

Longevity is an aspect that affects the tool's future, while reliability shifts attention to the
past and present. Since the user community is growing faster for free solutions, tools with
open-source code have a longer lifespan.

At the end of the chapter, after making an overview of the most popular tools, we will
weigh our defined criteria against their characteristics before selecting the RPA tool.

5.1 Selected tools

By that moment, we have explored multiple internet resources. Most influencing on

selection was two sources. The first one was PeerSpot, and we went through the Robotic
Process Automation section there (PeerSpot 2022b). This platform allows real users to
share their experiences using tools and technologies. The second influencer was a report
by Gartner, which is called "Magic Quadrant for Robotic Process Automation" (Gartner
2021), visual representation of it can be seen in figure 7.

25
Figure 7. Magic Quadrant for Robotic Process Automation (Gartner 2021)

In light of this, we have chosen the four most prevalent, popular, and promising
technologies available today.

5.2 UiPath

UiPath is a well-known RPA technology that allows users to automate manual, tedious,
and varied business processes. This open-source platform is generally used for
automating processes in PDF, online, Citrix, and Windows software applications
(PeerSpot 2022f). The UiPath has a huge community consisting of more than one million
users (Gartner 2021).

UiPath possesses several unique characteristics, including a comprehensive library of

activities, an automation center, process and task mining, and task recording. In addition,
it has server-side events to initiate automated procedures, database interaction, scraping,
PDF functionality, streamlined authentication, and marketplace templates (Gartner 2021).

26
Users may automate processes without previous programming experience. Trained AI
robots boost efficiency and productivity and improve overall customer happiness. With
UiPath Marketplace's drag-and-drop AI functionality and ready-to-use patterns, users can
automate more operations in less time.

It supports a wide range of operations and offers multiple kinds of robots (attended,
unattended, and hybrid) to accommodate an organization's demands despite the duties at
hand (PeerSpot 2022f).

UiPath is built of many components. UiPath Studio is one of them; it uses visual aids like
graphs and visualizations to develop automation. In addition, it applies the programming
and ordering of such procedures; its graphical user interface comprises pre-built
templates, interface buttons, and drag-and-drop capability. UiPath Robot is the
subsequent component, and these robots are placed to perform operations automation
according to established rules. UiPath Orchestrator is another component that works as a
web-based service for scheduling, deploying, and managing activities.

5.3 Microsoft Power Automate

Microsoft Power Automate is a universal software system that interfaces with a large
number of different services in a smooth manner. It was previously known as Power
Automate, Microsoft Flow, MS Power Automate. This solution could save costs on simple
activities utilizing Microsoft products. Microsoft charges if there are more than ten forms,
and there is no cost to maintain if less than ten forms are used. Based on it, most
companies can use this technology without paying anything extra. Office 365 licensing
includes Microsoft Power Automate as part of the general license (PeerSpot 2022e).

It is equipped with API integration and orchestration features. Tool's provider offers
comprehensive technologies inside the Power Platform, such as Power BI, Process
Advisor, Power Apps, API connections, and Power Virtual Agents (Gartner 2021).

This API-orchestrated solution can combine numerous record systems to automate data
transcribing tasks. Microsoft leverages all of the Power Platform components to build a
single, unified, end-to-end platform that provides automation, integration, low-code
application development, and analytics capabilities to fulfil companies' business process
automation needs. Power Automate offers a robust connection with Microsoft's ecosystem
of popular applications (Gartner 2021).

Microsoft’s RPA is dependent on Power Automate Desktop (PAD). It needs local

installation on computers or virtual machines operating on the most recent version of

27
Windows and cannot be installed on any other operating system. PAD is offered free in
Windows 10 and above versions. Also, there is a paid subscription that includes extra
features (Gartner 2021).

5.4 Blue Prism

Blue Prism is the pioneer of the RPA industry, and it was launched in 2001 (Taulli 2020).
Blue Prism is a clever, no-code automation tool for deployments of software as a service
in on-premises, public cloud, multi-cloud, and hybrid infrastructures. This tool, unlike
previous automation technologies, combines robotic process automation with increased
artificial intelligence and cognitive capabilities. It offers users with immediate access to the
tools necessary for constructing and deploying automation. Blue Prism brings better
flexibility and productivity by making it simple to automate the most important activities
(PeerSpot 2022c).

According to customer comments, Blue Prism excels in its enterprise-level capabilities.

Users mention that the solution is user-friendly, stable, and scalable in a robust manner. It
assists organizations with everything from compliance to quality control and mistake
prevention (PeerSpot 2022c).

A summary of this tool's features includes the following (Arora, 2021):

− Multi-environment deployment model.
− Automated credential management.
− Audit trails.
− Process changes in comparison.
− Process configuration.
− Application automation.
− Central administrative console.
− User role settings.
− Reporting and analytics.
− Integrations.

5.5 Automation Anywhere

Automation Anywhere is one of the most popular RPA solutions. Intelligent software bots
enable clients to automate end-to-end business operations. The firm provides a cloud-
native and web-based automation platform that combines RPA, artificial intelligence,
machine learning, and analytics with more security and scalability than traditional
monolithic solutions. Automation Anywhere has delivered almost three million bots to

28
serve some of the world's leading organizations in more than ninety countries (PeerSpot
2022a).

The following is a list of highlights for Automation Anywhere (Arora 2021):

− Cognitive robots.
− Rules setting.
− Machine Learning.
− Administrative console.
− Analytics.
− Single Sign-On.
− Version control.
− Bank-Grade Security.
− Integrations.

5.6 Decision Matrix for Robotic Process Automation tools

This section provides a table 1 comparing the investigated Robotic Process Automation
tools. In addition to the presence of criteria, we use the number of checkboxes within the
scope of our study to demonstrate the compatibility of technology with the business.

The table 1 presents eight items demonstrating the relationship and comparison between
four RPA solutions. We can observe that UiPath gained the most considerable number of
ticks compared to the other tools, while Microsoft Power Automate has the second-highest
score. For this comparison, we referred to users' feedback to these four tools on PeerSpot
(PeerSpot 2022d).

In general, all tools have equivalent technical features needed for our research, even
though UiPath has more options. One of the most important criteria was the cost of
ownership. Because it is included with our Microsoft Office subscription, Microsoft Power
Automate's main functionality is available at no additional cost, and in this phase, we do
not need extra features. UiPath also provides free options for individuals and small teams.
Based on feedback, the user experience for all these tools is fairly good, and some
reviews mentioned that UiPath is more intuitive. Exception handling, support and reliability
of the tools seem the same.

29
Table 1. Decision Matrix for Robotic Process Automation tools
Criteria UiPath Microsoft Blue Prism Automation
Power Anywhere
Automate

Technical
✓✓ ✓ ✓ ✓
features

Cost of
✓✓* ✓✓✓*  
ownership

Usability
✓✓ ✓ ✓ ✓
Exception
✓ ✓ ✓ ✓
handling

Support
✓ ✓ ✓ ✓
Reliability
✓ ✓ ✓ ✓
Community
✓✓✓ ✓ ✓ ✓✓
Longevity
✓✓✓ ✓✓ ✓✓ ✓✓

UiPath has the most sizable number of users and a big community, there are a lot of
forums, and it would be easier to find answers for common problems. All these tools have
a good longevity record and a promising future as leaders in the RPA market.

5.7 Summary of the chapter

No surveys tried to evaluate and consolidate these four RPA instruments. In this part, we
came to the point where we presented the most popular tools, and it is a moment to make
a selection of the tool among them to be used for implementing a manual process in later
chapters.

It is not easy to select since UiPath and Power Automate provide many valuable services
and essential features. Although UiPath has more ticks in a table, the final decision came

30
to Power Automate. Power Automate was preferred over other tools since it goes for free
as part of the current Office 365 subscription and allows us to save budget. This factor
was one of the most essential for us. This tool provides excellent usability, and it has all of
the features that we might need. Even though UiPath has the most significant community,
the Microsoft product community is growing and seem highly active. At the same time, we
have good support from the Microsoft side, which gives us an extra point to select Power
Automate.

Microsoft's product also seems quite promising, and the number of users is growing fast
every day, which bring us to the conclusion that it makes sense to invest efforts in this
solution.

31
6 The selected manual process

The chapter aims to provide an automation proposal for one of the manual invoice
handling processes in the finance and administration department. This chapter outlines
the series of actions done as part of a business process, the variables, and rules of the
process prior to automation, and how they are expected to function after delivering the
finished RPA process. The outcome of the chapter serves as a base for the subsequent
implementation chapter by giving the necessary specifications for deploying robotic
automation to the defined business process.

6.1 Summary of the process

The process selected for RPA is part of the daily manual routine that the finance and
administration clerks met in their work for organization "A".

The company "A" regularly provides many training sessions to their customers. It is
organized for customers worldwide, and each training can have from five to up to sixty
attendees. After such a session, the responsible person for the held session sends a
Microsoft Excel document to the accounting clerk. The document contains attendees'
information such as name, company, country, cost of the training, billing information, etc.

After receiving the document, accounting clerks process this file by manually filling the
invoice generation form with relevant data for each attendee by grouping per company in
fiscal management software called Procountor. Then, the clerk saves the filled invoice
form and clicks on send button for the invoice to be paid.

This process has reasonable grounds to be examined and eventually automated.

Successfully delivered automation will release accounting clerks from small routine tasks
and let them distribute their time for more critical duties.

6.1.1 Benefits and objectives

Implementation of the proposed manual process can bring different qualitative benefits to
the company:
− Increasing accounting clerk motivation by reducing a load of frustrating tasks.
− Allowing clerks to be more focused on important tasks.

At the same time, the clerk usually receives 1 to 3 emails per week with such excel
documents with training attendees. The actual handling of one attendee takes around 6

32
minutes per company. The finance and administration can achieve the following
quantitative objectives for savings:
− Automation can eliminate approximately 6 hours of routine work per week.

The main benefit of the actual implementation of RPA is that it opens gates for
researching and finding more business processes to be automated. After adding RPA to
the company as a widespread practice, even for a small task, it opens automation
possibilities for other employees and the idea generation process is triggered.

6.1.2 Implementation notes

The implementation effort and the size of the work are derived from key attributes. In the
table 2, it can be seen key attributes with their description and numbers which affect the
efforts.

Table 2. Attributes and descriptions of the implementation of the automation

# Attribute Description

1 Inputs The structure of receiving excel documents with attendees'

information differs based on the person who sent it. Although, the
accounting clerk can set the policy for training providers to bring such
information to a general template since the main needed fields for
invoice handling are the same. This template will be structured, and
the position of the details will be fixed. Inputs in fiscal management
software are always standard and structured and can be changed
only with a version update of Procountor.

2 Number of OS as starting screen – 1,

screens
Microsoft Outlook – 1,

Microsoft Excel – 2,

Procountor – 1,

Total – 4.

3 Type of Windows 10 – operating system,

Applications
Microsoft Outlook – email client,

33
 Microsoft Excel - spreadsheet software program,

Procountor – fiscal management software.

In real-business life, several roles are involved during the implementation of automation.
In this thesis, the author would wear most of the roles, and some of them by the
accounting clerk and thesis supervisor.

For clarity in descriptions of phases, it will be mentioned expected actual business roles in
brackets to be involved in automation because this thesis work can serve as a
constructive guide for implementing RPA in a company, even though only three persons
engaged in our study to implement the process.

First, implementation starts from a point where the author and supervisor of the thesis
(Business Analyst and Implementation manager) prepare robotic process automation by
defining business benefits and prioritizing the pipeline.

In the subsequent two phases, the author of the thesis takes the first phase and involves
an accounting clerk in the second phase for having access to Procountor. First, the
process goes to the solution design phase. The author (BA, QA, Solution Architect, and
RDA developers) describes how to implement a technical solution to the project and
prepares automation requirements. Next is the build RPA phase; in this step, the author
and accounting clerk (the RDA developer, PO, and BA) develop automation based on
previously written automation requirements.

In the test RPA phase, the author and accounting clerk (PO, RPA support, RDA
developer, and BA) work on performing user acceptance testing (UAT), bug fixing, and
testing.

After the previous step process moves to stabilize the RPA step, in this step, the process
goes to production (IT, RPA Developer, and Architect), then production monitoring by the
thesis author, supervisor, and accounting clerk (RPA support), and to measure the
expected results with actual results by business analyst (BA).

In the final phase, only an accounting clerk is involved. This phase ensures constant
improvement, where accounting clerks (BA, PM, and PO) perform performance
assessments. At the same step, benefits tracking can be done by the thesis supervisor
(PO) and managing changes can be performed by the author of the thesis (BA and the
Implementation team).

34
6.1.3 Background data

In the table 3, it can be seen the general data about the proposed automation process.

Table 3. Items and descriptions of the proposed automation process

# Item Description

1 Short name Invoice handling

2 Organizational level Finance and administration department

3 Input data An email with a special subject

4 Output data Filled invoice sent to be paid.

5 Period of execution Monday, 9 am – 10 am.

6 Average manual handling 6 minutes

time per item (AHT)

7 Number of items processed ~25

during a normal period

8 Peak period Randomly

9 Number of items processed >60

during peak period

10 The expected increase of Hard to estimate since the increase in the

volume in the future number of customers will lead to an
unexpected rise in the volume of attendees of
training sessions in the future

11 Total Full-time-equivalent ~0,15

(FTE)

35
6.2 Process overview

6.2.1 Applications used in the process

Table 4 includes a comprehensive list of all the applications that are used as part of the
process to be automated, at various steps in the flow.

Table 4. List of applications used in the process

# Application Language Access method Comments
name

1 MS Outlook EN Desktop Emails and

application data
management

2 MS Excel EN Desktop Spreadsheet

application software
program

3 Procountor FI Desktop Fiscal

application management
software

6.2.2 High-level AS-IS process map

This chapter depicts the As-Is business process at a high level to enable developers to
have a high-level understanding of the current process. Table 5 represents a simple high-
level process diagram.

36
Table 5. High-level AS-IS process map with average handle time in minutes
Open Read the email Download list of Open the file
Outlook with a list of attendees
attendees

Outlook ➔ Outlook ➔ Outlook ➔ Windows 10 ➔

0.5 1 0.2 0.3

Check the Open Fill the invoice Click the send

list of Procountor form for the invoice button
training attendee
attendees

➔ MS Excel ➔ Procountor ➔ Procountor ➔ Procountor

0.3 0.5 3 0.2

6.3 Detailed AS-IS process description

6.3.1 AS-IS process map

This chapter illustrates the AS-IS process flowchart. Figure 8 represents a workflow
diagram with data descriptions. A larger version of figure 8 is presented in Appendix 2.

37
Figure 8. Detailed AS-IS process map

6.3.2 AS-IS process steps

Table 6 describes the steps used in the workflow process map with defining input data,
description of the step, details, and output data.

Table 6. Numbered steps from the AS-IS process map

Step Input data Description of the step Output data

1 Open Outlook The user opens Outlook Outlook desktop

desktop version version was opened

2 Read an email The user opens the email The user has read the
with a request with a request to create email
invoices

3 Download the The user clicks on “Save all Excel document with
attachment of Attachments” button list of attendees was
the list of downloaded
attendees

4 Open an excel The user opens the The document was

document downloaded document open

38
5 Sort list of The user sorts the provided Document was sorted
attendees list of attendees based on based on company
based on company
company

6 Check The user reads the invoicing The relevant

invoicing details of the company company’s data was
details for the keeps it for later usage while read and kept.
company filling the invoice creation
form

7 Check training The user reads attendee The relevant

attendee from information, like name and attendee’s data was
this company keeps it for adding to the read and kept.
invoice creation form.

8 Open The user opens Procountor Procountor was open

Procountor desktop application

10 Open invoice The user navigates to the The invoice creation

creation form invoice creation form form was open

11 Fill company The user starts filling The invoice creation

invoicing invoicing details of the form was filled with
details company the company’s
invoicing details

12 Add to the filled The user adds attendee’s The invoice was filled
invoice the name to the filled invoice with the company’s
company and user’s details
attendee’s
name

13 Are there more A condition that checks Condition is checked.

training there are more training
attendees from attendees from this
this company? company, which verifies
another attendee's name
and adds to the invoice as in

39
 Step 12; otherwise, it goes
to the next step

14 Send the The user clicks on the send The invoice was sent,
invoice to the button for the invoice to be and the creation form
company paid was closed

15 Are there more A condition that checks if Condition is checked.

unprocessed there are no unprocessed
companies in companies then moves to
this document? the next step; otherwise, it
goes to the step 6

16 Switch to The user switches to Outlook was open

Outlook Outlook desktop application

17 Send an email The user writes a reply The employee

that the request email to the requestor that received the email
was processed the invoices were sent to the
companies of all attendees.

6.3.3 AS-IS data descriptions

The following table represents the data items used in the workflow process map. Each
data item is defined only once even though it is used as input data in many steps

Table 7. Data items from the workflow process map

Name of the Sample Input source Location Standard Structured
data (Yes/No) (Yes/No)

The email Standard Email Unopened Yes Yes

with a email with email in the
request the Outlook
document
file

Excel Standard Excel ‘\Downloads’ Yes Yes

document Excel sheet folder

40
 with a list of with a list of
attendees attendees

Procountor Standard Data from Installed Yes Yes

invoice Procountor excel desktop
creation invoice document application
form creation
form

6.4 Detailed TO-BE process description

6.4.1 TO-BE process map

This chapter illustrates the TO-BE process map in figure 9, and in Appendix 3, a larger
version of this figure can be found. We differentiate steps based on three colors. The blue
color represents only manual steps that will not be automated and will always be
performed by a human. In this phase of the research, it was decided to split the
automation scope and focus on delivering minimum viable product (MVP), which is why
we have green and red. As part of MVP, we would focus on implementing red-colored
steps to be automated and leave green colors for future iterations to be done as a future
improvement.

41
Figure 9. Detailed TO-BE process map

6.4.2 TO-BE process steps

The following table describes the steps used in the TO-BE process map with defining
input data, description of the step, details, and output data.

Table 8. Numbered steps from the TO-BE process map

Step Input data Description of the step Output data

1 Open Outlook The user opens Outlook desktop Outlook desktop

version version was
opened

2 Read an email The user opens the email with a The user has
with a request request to create invoices read the email

3 Download the The user clicks on “Save all Excel document

attachment of Attachments” button with list of
the list of attendees was
attendees downloaded

42
4 Open an excel RPA opens the downloaded The document
document document was open

5 Convert a RPA converts the document based on The document

document to a an accepted Procountor template was converted
specified
template for
Procountor

6 Open RPA opens Procountor desktop Procountor was

Procountor application open

7 Open invoices RPA navigates to the invoice The invoice

import form importing form import form was
open

8 Upload RPA uploads converted excel file to The document

converted excel Procountor was uploaded
file to
Procountor

9 Start importing RPA clicks on the import button Invoices were

invoices created based on
converted
documents and
placed in the draft
invoices tab

10 Open the draft The user opens the draft invoices tab The draft invoices
invoices tab tab was open

11 Are there A condition that checks if the draft Condition is

unprocessed invoices tab is empty then moves to checked.
draft invoices? the end of execution; otherwise, it
goes to the next step

12 Check invoicing The user reads the invoicing details of Company details
details for the the company and verifies it with an were verified
initial excel document

43
 unprocessed
company.

13 Check training The user verifies attendee information Attendee details

attendee from from the company with an initial excel were verified
this company document.

14 Are there more A condition that checks there are Condition is

training more training attendees from this checked.
attendees from company, which verifies another
this company? attendee's name; otherwise, it goes to
the next step

15 Send the invoice The user clicks on the send button for The invoice was
to the company the invoice to be paid sent, and the
draft form was
closed.

The flow goes to

step 11. If there
are no
unprocessed draft
invoices left. The
user moves to
MS Outlook.

16 Switch to The user switches to Outlook desktop Outlook was open

Outlook application

17 Send an email The user writes a reply email to the The requestor
that the request requestor that the invoices were sent employee
was processed to the companies of all attendees. received the
email

6.4.3 TO-BE data descriptions

The table 9 represents the data items used in the TO-BE process map. Each data item is
defined only once even though it is used as input data in many steps

44
Table 9. Data items from the workflow process map
Name of Sample Input Location Standard Structured
the data sourc (Yes/No) (Yes/No)
e

The email Standard Email Unopened email in the Yes Yes

with a email with Outlook
request the
document
file

Excel Standard Excel ‘\RPA\RawInvoices\’ Yes Yes

document Excel sheet folder
with a list of with a list of
attendees attendees

Converted Standard Excel ‘\RPA\ConvertedUnproce Yes Yes

document Excel sheet ssedInvoices\’ folder
for with a list of
uploading attendees
to
Procountor

Procountor Standard Data Installed desktop Yes Yes

draft Procountor from application
invoices draft conve
form invoices rted
form docu
ment

6.4.4 Business exception handling

In the detailed process map on figure 2, when the initial Excel file is converted to a
predefined Procountor template, several business exceptions can happen that should be
managed, in table 10 we listed these exceptions. The first exception can occur when
conversion takes longer than the defined timeout, then the process should stop and throw
an alert with a proper error message. The second exception can arise and throw the alert
if RPA met not valid fields in the initial file while converting to the Procountor uploading

45
file. The third exception can happen if uploading the converted file has failed based on a
predefined timeout, and then an alert should arise. Also, more business exceptions can be
added after the production at the “Constant Improvement” step.

Table 10. Numbered business exceptions of the process

BE# Step Description Input data Actions

1 5 Conversion of initial Data retrieved If conversion takes

document with list of from email with longer than the
attendees and all needed the request defined timeout, then
fields for invoicing based the process should
on Procountor template stop and throw an
alert with a proper
error message

2 5 Same description as in Same input data RPA should validate

BE 1 as in BE 1 fields in the initial
document while
converting to the
Procountor uploading
file; if validation fails,
RPA should stop and
throw an alert

3 8 and 9 Upload converted excel Converted file An alert should arise if

file to Procountor and uploading the
importing invoices converted file has
failed based on a
predefined timeout.

6.4.5 Reporting

In our case, reporting will be implemented as an alert pop-up with a successful message
that the RPA script was executed successfully. On table 11 it can be found reporting items
of the process.

46
Table 11. Reporting items of the process
R# Report name Frequency Description Report tool

1 Alert pop-up Each execution After each RPA tool built-

successful in functionality
execution, RPA
shows an alert
pop-up with a
message.

6.5 Acceptance testing

6.5.1 Normal test cases

Test case #1:

Description: Positive testing scenario.

Input data: We provide a dummy excel document with a list of training attendees where all
fields are valid.

Expected result: We manually go to the draft invoices tab and check that invoices were
created based on the data from the converted file and information in invoices corresponds
to data in the initial file.

Operating instructions: The human should prepare an input file with a list of dummy
attendees and then check the results manually.

Test case #2:

Description: Negative testing scenario.

Input data: We provide a dummy excel document with a list of training attendees where
some mandatory fields are empty.

Expected result: The program should tell us as a message that that nth row has the wrong
value in an input field.

47
Operating instructions: The human should just set the wrong input variables in the input
file with a list of dummy attendees and then check that alert with an error message will be
thrown.

6.5.2 Known business exception test cases

Test case #1:

BE#: Test business exception when the user provides a document with invalid input fields.

Input data: We provide a dummy excel document with a list of training attendees where
some mandatory fields are empty.

Expected result: The program should tell us as a message that that nth row has the wrong
value in an input field

Operating instructions: The human should just set the wrong input variables in the input
file with a list of dummy attendees and then check that alert with an error message will be
thrown.

6.6 Summary of the chapter

In this chapter of the thesis, it was selected the manual process, and we made an
automation proposal for one of the manual invoice handling processes. We defined that
we would focus on delivering the minimum viable product by selecting the most crucial
steps for automation. We wrote a detailed process definitions document which will be
used as a guide in the subsequent chapter of the RPA implementation phase.

48
7 Automation of the process

The chapter introduces the actual use of Power Automate Desktop. It describes general
actions used in the implementation and shows in which use cases the implementor used
these actions. Besides, that chapter gives a profound explanation of the implemented
flow. The outcome of the chapter is made and delivered RPA flow to the accounting clerk.
Also, it serves as the last chapter before the conclusion.

7.1 Introduction of Power Automate Desktop

Microsoft Power Automate allows building automation from start to finish using Power
Automate flows. Building desktop flows is the first step in designing an RPA solution.

First, we go to Power Automate and sign in with the Microsoft account. After successfully
signing in, we should select "My flows", "Install", and "Power Automate for Desktop", like it
is shown in figure 10. It downloads the Power Automate for Desktop installation file.

Figure 10. Manage your flows in Microsoft Power Automate

After downloading, we select the file and start the installation. In figure 10, after clicking
the "Next" button on several installation screens and selecting the last checkbox, we
install the tool. Besides installing Power Automate, it installs browser extensions for
Google Chrome and Microsoft Edge for web-automation steps.

49
Figure 11. Install Power Automate package

After successful installation in figure 12, select the link of your preferred browser in the
appeared window and enable the Power Automate extension in a browser. Then Power
Automate desktop is ready to be launched.

Figure 12. Installation successful of Power Automate Desktop

50
By that moment, we had downloaded the right software and had everything to create the
first desktop flow. At this point, the primary duty is to become acquainted with the Power
Automate Desktop.

First, we open and start exploring the app in figure 13. Under Flows, we can see created
desktop flows, which have the functionality to edit, start, or delete these flows. We can
explore how to create or edit a flow by clicking the "+ New flow" button in the upper-left
corner.

Let us name this new flow to be "Training attendees invoicing". Then we click on the
"Create" button.

Figure 13. The main view in Power Automate Desktop

After creating a new flow, the Power Automate designer view will appear, like it is shown
on figure 14. The view has several elements to be concentrated on:
− The toolbar contains basic actions such as "Save", "Undo", "Copy", "Paste”, and
"Debug". It also includes buttons to start the desktop and web recorders and control the
implementation process by "Start", "Pause", or "Stop" buttons.
− "Subflows" button allows us to create subflows under the Main flow, which can be
mentioned as a group within a desktop flow, similar to private methods in programming.
− The actions pane includes all Power Automate actions and a search bar that assists in
finding specific actions by matching the action name to the typed words.
− The workspace contains all actions that have been added while implementing the
automation process. These procedures are separated into tabs.

51
− The "Input/output variables" panel holds variables used in data exchange between
cloud and desktop flows.
− The "Flow variables" panel lists all variables we created while implementing the
automated process.

Figure 14. Power Automate designer menu

After clicking the "Save" button and closing the Flow designer window, we see a newly
created flow with the name "Training attendees invoicing" in the list in figure 15. The
accounting clerk will use this view for executing the developed RPA process by clicking
the "Run" icon.

52
Figure 15. Main view with created flow in Power Automate Desktop

7.2 General actions

While the development phase, we used many actions provided by Power Automate
Desktop. We used each of them many times to accomplish different use cases. In this
subchapter, we would like to describe each action in order of appearance with one
example of the use case without repetition of the same action, even if we used it for many
other use cases.

The first action used in the flow is "Display custom form", which lies under the "Message
boxes" category. By using this action, we can create a custom form shown in figure 16.
This action has its custom form designer, which contains all different input elements. We
can add text, date, and file inputs to our custom form. For making the form, we use a
simple drag and drop technique. The inputted data is saved as one custom form object
from which we can retrieve and use the information in our flow. In the first occurrence of
this action, we created a form where we provide input files for fetching training and
companies’ data and radio buttons for specifying quantities of training attendees' files to
be processed.

53
Figure 16. Custom form designer

Then, "Display message" and "Display input dialog" are used several times. Primary use
cases were to notify the user about the flow state, like finishing or termination. When
some data was missing, it was asked to input needed data in this dialog manually.

Once, we used the "Display select file dialog" when we selected the files, we wanted to
transform into the upload file for creating draft invoices.

Next, commonly used actions are part of the conditionals category. We used all of them in
the implemented flow. It includes "Case", "Default case", "Else", "Else if", "If" and "Switch".
They allow us to execute blocks of actions only if a given condition is met. Example of “If”
condition designer is shown in figure 17. If the condition is false, the automation flow will
skip the block of actions. Using them, we can ensure certain elements have a wished
value or state before performing other actions.

54
Figure 17. If condition designer

For example, in our flow, after showing the custom form, we check the produced variable
of which button was pressed, and if the pressed button equals done, we execute actions
inside the "If" block shown in figure 18.

Figure 18. If condition block

Another commonly used action is "Run subflow". We used it to execute previously created
subflows to split the code and make it more readable.

55
Within flows, variables are utilized to store data for future processing. Each variable's
name must be surrounded by per cent marks (%) like it is represented in figure 19. The
percentage symbol is utilized to express variables as a special character.

Figure 19. Set variable window

Power Automate permits the construction of complicated expressions with hardcoded

variables, lists, variable names, arithmetic and logical procedures, comparisons, and
parentheses.

Another essential part of the flow is Excel actions. After setting an Excel instance with the
Launch Excel action, we read from an Excel file. We use the Read from Excel worksheet
action to read and extract data from an Excel document. Retrieve the value of a single cell
or a table of data. When the flow runs, the action stores the data in a data table type
variable.

56
Figure 20. Excel actions

In figure 20, flow opens excel, reads data from the opened file, saves it in the data table
with the name TrainingCodesData and closes the Excel instance.

Another important actions worth mentioning are loops. They automate repetitive sections
of a flow by executing a block of actions multiple times. Loops are a fundamental concept
in desktop flow development and are invaluable in complex flows. The main idea behind a
loop is to make a desktop flow repeat one or more actions multiple times. Power
Automate includes three separate loops that cycle according to various conditions; we
utilized them in our workflow. Simple loops run a predetermined number of times.
Conditional iteration when a condition is valid. For each iteration, we iterate over a list.

In figure 21, we iterate through selected files for transforming and performing predefined
operations with each selected file.

Figure 21. For each loop

While working on automation, was met a problem with the slow handling of companies
while fetching company codes. Because the file with these data contains over two
thousand rows of records, we had to start using Database actions instead of loops for
reading companies' data to address this problem and improve the flow performance.
These actions include an opening, executing, and closing SQL connection.

57
By introducing these actions, while transforming ten lists of attendees with twenty
attendees in each file, the execution time was fifteen times faster than implementation
with loops for finding company codes.

7.3 RPA process overview

In this subchapter, we would like to focus more on a process overview from a user
interaction perspective than a code development site. The actual code and solutions
applied in the script development can be observed in Appendix 1.

The first step for triggering the implemented RPA process is to open Power automate.
This window in figure 22 presents all implemented RPA flows from which we can run or
edit them.

Figure 22. Main view in Power Automate

Next step, we click on the run button and the flow starts. Power Automate notifies us in
right bottom corner in figure 23 that the flow has started, and we can pause or stop the
process through this window.

58
Figure 23. Power Automate notification window

The first step in the flow is to select the amount of files, training codes and companies'
codes which are shown in figure 24. We must choose the number of files in data
preparation since the flow adds training and company codes for each file. In our example,
let us select multiple files.

Figure 24. Training attendees invoicing first window.

Next, a file selection upload pop-up appears in figure 25, and it asks us to select multiple
files.

59
Figure 25. Files selection

Next, it asks us for each selected file to specify which type of training it was in figure 26,
and if the flow could not recognize by itself the training date and country, it would leave
empty fields for it, and we will need to fill it.

Figure 26. Date, country, and training type selection window

If the flow meets problems while finding company code, it will show the window selecting
the closest company code or suggest typing company code manually for a troubling
company in figure 27.

60
Figure 27. Company code exception handling

In the end, the flow will notify us that script was successfully executed in figure 28. After it,
we can go to the predefined output folder and use the generated transformed file with all
attendees for uploading to Procountor and creating draft invoices for sending to
customers.

Figure 28. Success notification window

7.4 Summary of the chapter

In the implementation phase, we met several problems. The first problem was that the
prepared documentation missed several specifications uncovered while developing the
flow and was written without proper knowledge of Power Automate and development
efforts. It was one of the reasons we decided to deliver a minimum viable product and left
only part with transforming raw data to uploading document file based on the Procountor.
An accounting clerk can manually upload since it takes one minute to do this. At the same
time, the actual automation of uploading can delay the delivery of the flow, and we can do
it in the following development iterations.

The second problem was that not all scenarios of exception points were taken while
working on the process definition document. For example, we have not thought that there
were two thousand companies and codes for them, and this information is missing in an
initial file. We need to fetch it from separate sources. For example, the name of a
company used in the default file cannot be fully matched to one in the company's file or
having several companies with the same name but for different markets. We also had to

61
spend the effort optimizing script and working on performance while fetching companies'
codes and using queries to oversee this problem.

There is more scope for further improvements, but the main part was done and can be
used by the clerk daily. Another crucial point is that we set the RPA practice in a
company. Now we have ready-made solutions that can be applied to implementing new
flows to automate manual routines. We did the most challenging part, and the further
process will be more straightforward with the next RPA scripts.

62
8 Discussion

As stated in the first chapter, this thesis aims to trigger the use of RPA within the
organization for handling manual tasks. To reach this goal, we surveyed manual
processes in the accounting department for automation. We created detailed development
specifications with test cases; after that, we researched RPA tools on the market and
selected the tool to be used. We automated the selected manual process based on written
specs, and as the last step, we delivered an automated flow to the accounting
department.

After reading a thesis, we expect the survey results to help readers comprehend the steps
we took to establish RPA practice in a company. We feel that we researched the latest
RPA tools on the market and utilized the best available sources.

The thesis gives clear answers for the first two sub-questions. Nevertheless, we cannot
answer straight about the effects of RPA on the work process after implementation
because it will take time before the process will be fully used and adjusted. Still, we can
easily ensure that by that moment, time spent by the clerk is reduced from ten hours to
one hour for overseeing the automated selected manual process.

In a thesis, if we consider the reliability of our findings, it is worth mentioning that we used
constructive research with qualitative and quantitative research techniques. When we tried
to find the manual process for robotic process automation, we relied on qualitative
research methods by collecting the data from the most involved and aware respondents
about finance and accounting processes in a company. It was used several interview
methods. First, we started with the semi-structured interview, then used an open-ended
interview, then a focus group interview and finally, we had a brainstorming session.
Although, when we moved to the RPA tools chapter, we relied on dozens of reviews and
feedback from the RPA community to define the best RPA tool, so this part we
accomplished with the help of quantitative research methods. The results of the thesis
work can be considered trustworthy because if similar research is conducted under
identical conditions and for a comparative accounting process and the company with the
same Microsoft subscription, the results will be the same.

Regarding the validity of the results, we can state that our findings are accurate to the
current company situation and the methods used in a thesis with data were reliable
because of collecting data from trustful sources.

63
Considering the ethical viewpoints, we can state that in the research, we tried to pursue
high moral standards and do the work honestly and carefully, aiming for the results to be
helpful. During the research process, we explained to all involved people the actual goals
of the research and their roles in the study. We kept all answers anonymous and emerged
conclusions based on interviewed persons will. Also, we can ensure that we did not falsify
or mispresent the results of the study to make the desired or polished findings and
conclusions. We did our best by mentioning and citing all articles, web sources or
someone's texts in the thesis.

This thesis has demonstrated that further research in this area has considerable potential.
During research for a thesis, countless ideas evolved. For example, in subsequent
iterations of RPA improvements, we can finish Procountor uploading side automation;
then, as part of the next iteration, we can automate manual validation of draft invoices
created after uploading the file.

As an author, I can say that the thesis process was challenging. Even though it went
smoothly, I always had thoughts and ideas about what to write. I had continuous support
and interest in the topic in all phases of the thesis process from my thesis supervisors and
all involved persons. I felt burned out before starting a thesis in the summer of 2021 when
I completed most of the studies within seven months with a full-time job at the same time,
so it was hard to push myself to write the thesis at the same pace as I passed courses in
the MBA program. But slowly and step by step, I was able to finish the thesis work but
failed the initial very optimistic planned schedule to complete a thesis within five months.

In the early phases of the thesis, it was hard to define the write scope of the objectives
and main research question because, at that moment, I had a broad thought without
having clear expectations and goals for a thesis. Although I successfully set them and
followed research questions by trying to answer them throughout the whole thesis
process. In the next stage of the thesis process, it helped me a lot that I covered the
applied design research course, which uncovered the proper research methods which
should be used in this study and showed me how to conduct the research.

While working on a thesis, I could apply the knowledge I gained in several courses in the
MBA program. I also surveyed several RPA tools and learned how to write automation in
Microsoft Power Automate. There is a belief that no coding is required to start working
with RPA. However, my programming background has helped me a lot and reduced a
time in the learning curve while gaining this knowledge because the same patterns and
logic are applied while working on RPA.

64
In conclusion, this survey provides a deep dive into RPA technology; therefore, it might be
valuable for people seeking a guide that complements a foundational understanding of
automation for getting started. This research enables the reader to get familiar with what
RPA is, how to define manual tasks, meet with the newest popular tools, and briefly
introduce the coding of the RPA flow without wasting time on redundant information.

65
Sources

Arora, S. 2021. A Robotic Process Automation Tools Comparison. URL:

https://www.simplilearn.com/robotic-process-automation-tools-comparison-article.
Accessed: 29 August 2022.

CiGen 2021. 7 Challenges to Implementing Robotic Process Automation (RPA) & How to
Overcome Them. URL: https://www.cigen.com.au/7-challenges-implementing-rpa-how-
overcome-them/. Accessed: 29 August 2022.

De Bono, E. 2017. Six thinking hats. Penguin u.

ElectroNeek, 2021. RPA for Finance & Accounting: 10 Best Use Cases. URL:
https://electroneek.com/blog/business-insights/rpa-for-finance-accounting-10-best-use-
cases/. Accessed: 29 August 2022.

Ernst & Young 2021. Robotic process automation in the Finance function of the future.
URL:
https://insightsbenelux.com/publications/%2314_werkplaats_(ochtend)_EY_Finance_Rob
otics.pdf. Accessed: 29 August 2022.

Gartner 2021. Magic Quadrant for Robotic Process Automation. URL:

https://www.gartner.com/doc/reprints?id=1-
26Q65VFT&ct=210706&st=sb&mkt_tok=OTk1LVhMVC04ODYAAAGBbAd50L4FRSBds9
6Gdo8NUnY4J0_xxXbIINK7peZ65aWh03yVDGB6BatGuSrQuYGLDaQJaRp371AzV2nyQ
c_dAKCObPjRwdEAZGxDJ7v2VS6b. Accessed: 29 August 2022.

Harmoko, H. 2021. The Five Dimensions of Digital Technology Assessment with the
Focus on Robotic Process Automation (RPA). Tehnički glasnik.

Kedziora, D., Leivonen, A., Piotrowizc, W. & Öörni, A. 2021. Robotic Process Automation
(RPA) Implementation Drivers: Evidence of Selected Nordic Companies. Issues in
Information Systems.

Howe., J. 2006. The rise of crowdsourcing. Wired.

Ojasalo, K., Moilanen, T. & Ritalahti, J. 2021. Methods for Development Work. Helsinki.

Osborn, A. F. 1953. Applied imagination.

66
PeerSpot 2022a. Automation Anywhere (AA) Reviews. URL:
https://www.peerspot.com/products/automation-anywhere-aa-reviews. Accessed: 29
August 2022.

PeerSpot 2022b. Best Robotic Process Automation (RPA) Tools. URL:

https://www.peerspot.com/categories/robotic-process-automation-rpa. Accessed: 29
August 2022.

PeerSpot 2022c. Blue Prism Reviews. URL: https://www.peerspot.com/products/blue-

prism-reviews. Accessed: 29 August 2022.

PeerSpot 2022d. Comprasion of RPA leaders. URL:

https://www.peerspot.com/products/comparisons/automation-anywhere-aa_vs_blue-
prism_vs_microsoft-power-automate_vs_uipath. Accessed: 29 August 2022.

PeerSpot 2022e. Microsoft Power Automate Reviews. URL:

https://www.peerspot.com/products/microsoft-power-automate-reviews. Accessed: 29
August 2022.

PeerSpot 2022f. UiPath Reviews. URL: https://www.peerspot.com/products/uipath-

reviews. Accessed: 29 August 2022.

Pritzker, M. 2020. RPA Drawbacks and Limitations Explained 2020. URL:

https://www.peerspot.com/questions/what-are-the-limitations-of-rpa. Accessed: 29 August
2022.

Sutton, R. I. 1996. Brainstorming groups in context: Effectiveness in a product design firm.

Administrative science quarterly.

Taulli, T. 2020. The robotic process automation handbook.

UiPath 2021. How robotic process automation (RPA) will transform F&A. URL:
https://www.uipath.com/resources/automation-whitepapers/rpa-finance-accounting.
Accessed: 29 August 2022.

67
Appendices

Appendix 1. Extract of implemented main RPA flow source code

@@statistics_TextBlock: '1'

@@statistics_Input_ChoiceSet: '1'

@@statistics_Input_File: '2'

@@statistics_Action_Submit: '2'

Display.ShowCustomDialog CardTemplateJson: '''{

\"type\": \"AdaptiveCard\",

\"version\": \"1.4\",

\"id\": \"AdaptiveCard\",

\"body\": [

\"type\": \"TextBlock\",

\"id\": \"Text block\",

\"color\": \"dark\",

\"italic\": true,

\"text\": \"${Text_block_Text}\",

\"horizontalAlignment\": \"center\",

\"maxLines\": 3

},

\"type\": \"Input.ChoiceSet\",

\"id\": \"Choice set input\",

\"style\": \"expanded\",

\"isMultiSelect\": false,

\"choices\": [

\"title\": \"One file\",

\"value\": \"One file\"

},

\"title\": \"Multiple files\",

\"value\": \"Multiple files\"

68
Appendix 2. Detailed AS-IS process map

69
Appendix 3. Detailed TO-BE process map

70