This is the failure that will affect you and that almost nobody designs for. An approval step only works while the person is genuinely reading it. Ask someone to approve forty similar things a day and by week three they are clicking approve without reading, and your safeguard has quietly become a button that adds latency.

Causes Fatigue	Preserves Attention
Approving everything, including safe things	Only irreversible actions need a yes
Forty approvals a day	Under ten a day
Every request looks identical	Unusual requests look visibly different
No consequence for approving wrongly	Occasional sampling and feedback on decisions
Approval is a modal blocking their work	Approval is batched at a natural moment

Measure Your Approval Rate

If 98% of requests are approved, the approval step has probably stopped being a decision. That's not automatically wrong, it may mean the agent is reliable, but it means you're getting no safety from the step. Either reduce what requires approval so the remaining ones get real attention, or accept that you're running at a higher autonomy level than you think.

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

Approval Fatigue