AI Agents

Summary

AI Agents is a focused course on agentic AI: systems that break a goal into steps, use tools along the way, and act rather than only answer once. It builds directly on the Beyond Chat material in AI Foundations and goes deep into design, oversight and the specific failure modes that come with systems that run for longer than a single reply.

Who Should Attend

• Technical leads, process owners and innovation or IT staff evaluating or piloting agentic AI systems
• Teams already using AI Automation who are considering the next step toward more autonomous processes
• Anyone who needs to assess vendor claims about ‘AI agents’ accurately, rather than take them at face value

Course Description

The course opens by drawing a firm line between an agent, a chatbot and a simple automation, because vendors routinely blur it. An agent plans its own sequence of steps, decides what to do next based on what happened at the last step, and can run for minutes or hours rather than seconds which is what makes it useful, and what makes it risky.

Most of the day is spent on the four controls that make an agentic system usable in practice: human approval at the point of consequence, a visible and reviewable trail of every step, a narrow scope before a broad one, and a defined stopping condition rather than open-ended improvisation. This is the control-tower pattern introduced in AI Foundations monitor, investigate, recommend, approve, execute, verify expanded here into a full design method participants apply to a real candidate process.

The course closes on governance and accountability specifically for autonomous systems: because the process ran itself does not change who is answerable for its output, and the course is direct about what that means in practice.

Duration and Format

• Length: One full day, a mix of conceptual grounding and applied design work
• Format: Hands-on where agent-building tools are available; otherwise worked through as guided design exercises
• Audience note: Technical staff will get more from the applied sections, but no coding is required to complete the course

Pre-Requisite

Participants should:

• Have completed AI Foundations, in particular the Beyond Chat module
• Completion of AI Automation is recommended, though not required
• Bring a real process that is a candidate even a rough one for agentic or highly automated handling

Course Outline

     01: What an agent is, and isn’t
     02: Tool use and grounding
     03: Designing for failure
     04: Evaluating and piloting an agent for a real process
     05: Governance and accountability for autonomous systems

What You Will Be Able to Do

• Explain accurately what distinguishes an agent from a chatbot or a simple automation
• Evaluate vendor claims about agentic products against a working definition, not marketing language
• Design the four controls that make an agentic system usable: approval, audit trail, scope, stopping condition
• Scope a safe, narrow pilot for a real agentic use case in your organisation
• Apply the control-tower pattern monitor, investigate, recommend, approve, execute, verify to a real process

You Leave With

• A scoped pilot design, with controls specified, for one real agentic use case
• A control checklist approval points, audit trail, scope, stopping condition for any future agent project
• A set of specific questions to take into your next vendor demonstration