The Decide step asks: am I done? Getting this wrong produces the two most common agent failures, and the decision about where to draw these lines is yours, not engineering's.

Stops Too Early	Never Stops
Answers before it has the facts	Loops on the same action repeatedly
Gives a partial answer confidently	Keeps investigating past the point of value
Fails to notice its tool call failed	Burns budget on a task it can't complete
Caused by a weak goal definition	Caused by no cap and no give-up condition

Every Agent Needs a Give-Up Condition

Not a crash, a deliberate, designed surrender. 'After 15 steps, or 90 seconds, or 0.40 pounds, stop and hand to a human with everything found so far.' Without one, an agent that gets stuck will keep going until something else stops it, and that something is usually a timeout that discards all the work. Design the surrender; it's a product decision with a user-facing outcome.

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

Stopping Conditions