IBM’s vision
for agentic AI
The future of AI is about
agency and productivity
Matt Denham
[email protected]
Learning Content Development, Data Platform
AI capabilities are
growing rapidly
AI that creates AI that predicts AI that chats AI that works
Generative AI Machine learning AI assistants AI agents
With agentic AI,
today’s front-end
systems will become
back-end systems.
AI agents will drive the next wave
of knowledge worker productivity
25% 12% 40%
decrease in time to complete more tasks completed in total higher quality in work
tasks when using AI tools1 with the help of AI tools1 completed with AI tools1
By 2028, 1/3 of interactions with generative AI will
use autonomous agents for task completion2
What is an AI agent?
An AI agent is an application
that can act autonomously to
understand, plan, and execute
a specific task.
AI agents use LLMs to reason
and can interface with tools,
other models, and other IT
systems to fulfill user goals.
What is agentic AI?
Agentic AI is a framework
for accomplishing goals
with limited supervision
that consists of AI agents.
In multiagent systems, each
agent performs a specific subtask
that’s required to reach the goal.
Agentic AI across industries
Healthcare Education Gaming Insurance
Multi-agent systems Autonomous agents can Agents which observe Agents can have a
can be useful for offer tailored learning and interact with human dramatic impact on
problem-solving, such plans and adapt players can provide resolution time with
as treatment planning teaching methods realistic experiences claims, policies, etc
Use case
Reshaping HR with AI agents
Benefits
1. Pre-built integrations designed to seamlessly integrate
with existing HCM platforms, ticketing tools, and more.
2. Pre-build skills which allow users to quickly start
testing a range of automations from requesting time
off to initiating employee transfers.
3. Simple and intuitive customization means HR teams
can configure the system to enterprise requirements
and policies
“You can train it do certain tasks within one process, but
you can easily have it apply those same skills to other
processes. It blows chatbots out of the water. It really
is changing our understanding of the future of work.”
Jon Lester, VP of HR Technology, IBM
Read the blog
HR agent
demonstration
9
Shadow AI is an emerging threat
38%
of workers share sensitive work
69%
of workers say they are unaware if their
information with AI tools without companies have formal strategies or
their employers' permission1 plans for integrating generative AI2
Be proactive! Provide knowledge workers
with governed and secure agentic AI tools.
Best practices for deploying AI agents at scale
How can an enterprise manage 100s of autonomous agents?
Flexible tools for Management and Security Guardrails and
developing AI agents governance explainability
• Low-code tools for • Lifecycle management • Manage authentication • Set conditions for
knowledge workers • Collect agent decision across systems human intervention
• Development frameworks telemetry data • Secure data exchange • Guardrails to ensure
and APIs for coders • Monitoring and alerts between tools and services agents meet
• Monitor for attacks and expectations
vulnerabilities and • All agent actions and
decisions to be
explainable
IBM’s architecture
Customers Employees Knowledge workers
for a scalable and
responsible agentic
Orchestrator agent
AI practice AI agent supervises and manages how work is
Supervise Route Reason Plan
executed across agents, assistants, and skills
Skills AI assistants AI agents
Intelligent task, decision, document and AI-assisted experiences that are Autonomous AI-driven execution
workflow automations augmented by gen AI human trained and designed of expert tasks
Agent building studio
Discover, create and manage gen AI and digital automations that combine decisions, tasks, skills, and workflows
Prebuilt skills and app catalog Custom skills – build Gen AI catalog Agent catalog
new, discover existing powered by
leading models Agents for Agents for
HR Procurement
Agents for Custom-built
Sales Agents
Management Developer toolkit
Common tools for administration, AI governance, AI security Tools, open-source frameworks, templates, deployments
Hybrid cloud platform
Start quickly with a
domain-specific AI agent
Preview 1H 2025 1H 2025 1H 2025
watsonx Agents watsonx Agents watsonx Agents watsonx Agents
Choose a domain specific AI for HR for Sales for Procurement for Customer Care
Agent that comes packaged with Transform HR functions Boost sales Augment agility via Transform CX &
prebuilt skills to accelerate the & employee self-serve performance & AI-driven digitization enable customers to
augment buyer
journey to launching an experience
of tasks self-serve
enterprise grade AI Agent • Employee support • Sales enablement • Supplier • Customer self-serve
• Talent acquisition • Prospecting support assessment (w/ • Modernize call center
• Expense Management • Opportunity mgmt.. D&B data) • Contact center insights
• Source to pay • Agent assist
• Procurement insights
Starting with HR, leaning on IBM’s over 7 years of experience
Agents EMPLOYEE SUPPORT TALENT ACQUISITION ONBOARDING
preview Coming in 2025 Coming in 2025
ROI
40% 94% +55 98% 30% 10% 88% 10% 5%
reduction in HR employee inquiries improvement in HR NPS faster candidate increase in hires per increase in employee reduction in onboard decrease in errors from reduction in
operating budget1 handled by AskHR1 score (+19 to +74)1 communications2 forms completion4 manual work5 employee turnover5
month2 happiness3
• Time off management • Create job requisition • Initiate pre-employment checks
• Employee management • Find candidates for job requisition • Recommend learning for knowledge gaps
• Profile management • Schedule interview for a candidate • Initiate employee onboarding
Flows • Support • Interview transcript summarization profile creation
• Employee verification • Email rejection letter to candidate • Create IT requests
• Benefits • Create offer packages
• Compensation
Apps
Accelerate the time-to-value of AI agents with IBM
Build intelligent Provide
interactions end-user
Conduct Q&A Integrate and automate processes access
Built-from- Intent testing
scratch Manually build flows Connect to enterprise apps & refinement Custom
API function mapping
solution Process automation across apps
chat
IT review
80% less effort 70% less effort 75% less effort 90% less effort
Connect to policy docs Leverage pre-built With agentic routing Pre-built AI chat
instantly customizable skills
IBM
solution
70%
reduction in time
to go live1
1Based off an internal study using data from IBM AskHR implementation
Time (months)
IBM Business Partner and internal use only
Custom-built agents
watsonx.ai allows developers to build, deploy and
manage the end to end Agent Lifecycle
Agent Lifecycle POV
Agent Creation Phase Agent Management Phase
Plan Build Agent Runtime Monitor
Deployment Orchestration
• Model Support • Agent Creation • Agent onboarding • Runtime • Monitoring and
• Tool Support • Templates and verification recommendations, alerts
• Customization • Deployment allocation, and • Auditability
orchestration
Governance Evaluation Monitoring Observability Customization
• Auditability • Build-time Agent Eval • Alerting • Tracing • Tuning Studio
• Runtime agent eval • Instruct Lab
• Eval of trace data • Synthetic Data Generation
for Agents/Tool Calling
• Tool Customization
AI use cases Customer experience Digital labor IT operations
that unlock
distinct value
Customer service Human resources Application modernization
33% increase in Reduced time from résumé Gained 30% productivity3
agent efficiency1 submission to offer by 75%2
Customer/citizen service Procurement Application modernization
800 hours saved Over USD 40 million in cost 1,500 hours of manual
each month4 savings and avoidance5 labor saved yearly6
USD 3 billion saved Marketing Supply chain IT operations
67% content creation Reduced supply chain 70% of Ansible Playbook
efficiency gain for costs by USD 150 million8 content created by watsonx
supported asset types7 Code Assistant9
Four ways to start your
AI agent journey with IBM
Free trial Client briefing Agentic AI bootcamp Pilot program
Test out watsonx to Discussion and custom Learn top Agentic AI skills Engagement with a multi-
build AI models, demonstration of IBM’s gen and best practices resulting disciplinary IBM team to
customize agents, and AI capabilities. Understand in a functional prototype jointly innovate and prove
accessing data across where gen AI can be used which addresses an the business value of
your organization. now for impact in your enterprise use case. generative AI solutions
business. using watsonx.
2-4 hours 1-2 days 1-4 weeks
Link → Link → Link → Success story →
