A small error at step two becomes the foundation for steps three through twelve. By the end the agent is confidently building on something wrong, and the output looks coherent because every step after the error was internally consistent.

Step	What Happened	Consequence
2	Misread the shipment reference, 44182 as 44128	Now investigating the wrong shipment
3-7	Correctly investigated 44128	All findings are internally consistent and irrelevant
8	Drafted an accurate update about 44128	Reads perfectly
9	Ready to send to the customer of 44182	A confident, well-written, completely wrong message

Nothing about the output signals the problem. The defence is checkpoints: having the agent restate key facts at the point of action. 'Sending an update about shipment 44128 to the contact for 44128' would have surfaced the mismatch immediately, at the one moment a human was looking.

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

Compounding Errors