A large share of products marketed as agents are workflows with an LLM inside. That's not a criticism, workflows are often the right choice, and Chapter 9 argues for them. But calling a workflow an agent causes real problems, because it sets expectations about flexibility the system doesn't have.

It's a Workflow If	It's an Agent If
You can draw the steps on a whiteboard	The steps depend on what it finds
The same input always follows the same path	The same input might take four steps or eleven
You know the cost per run within 10%	Cost per run varies by a factor of five
Adding a capability means adding a branch	Adding a capability means adding a tool
A failure means a step broke	A failure might mean it chose badly

Use the language honestly inside your team even if marketing uses it loosely outside. If your engineers think they're building an agent and you think you're specifying a workflow, you will discover the mismatch late and expensively.

Agentic AI and MCP for Product Managers

Module 1What 'Agentic' Actually Means

Module 2How an Agent Actually Works

Module 3MCP: What It Is and Why You Should Care

Module 4Designing the Tool Layer

Module 5Permissions and the Human in the Loop

Module 6Evaluating Agents

Module 7Cost, Latency and Unpredictability

Module 8Designing What the User Sees

Module 9Risk and Blast Radius

Module 10When an Agent Is the Wrong Answer

Module 11Final Project: Design One Real Agent End to End

Module 12Appendix: The Prompt Library

Why Most Things Called Agents Aren't