What an Agent Actually Is, and What It Cannot Do

An AI agent pursues a goal through multiple self-directed steps — planning what to do, calling tools such as databases or APIs, and tracking what has already happened across the session. A chatbot receives a single prompt and produces a single reply: one turn, done. This course teaches you the three structural markers of genuine agentic behaviour (planning, tool use, and state-tracking) so you can tell the difference when a vendor makes a claim.

This course is designed for non-technical business leaders — including board members, founders, product managers, and senior executives — who need to evaluate agent deployments, interrogate vendor claims, and decide when a task is appropriate to hand to an autonomous AI system. No coding or machine learning background is required. The focus is governance and critical evaluation, not technical implementation.

No prior AI or technical knowledge is needed. Familiarity with everyday business software is helpful but not required. The course starts from first principles and builds the vocabulary you need to evaluate agentic AI from a leadership and governance perspective.

The course totals 35 minutes of narrated content across 4 modules and 8 lessons. Individual lessons range from 3.5 to 5 minutes, so the material can be completed in a single sitting or spread across short sessions.

After completing this course you will be able to distinguish a genuine agent from a rebranded chatbot using a concrete checklist, recognise the four dominant production failure modes of agents, critically evaluate vendor ROI and benchmark claims using real disputed cases as reference points, determine whether a proposed deployment meets minimum governance conditions, and draft three targeted due-diligence questions to ask a vendor or internal team before approving a deployment.