01 Separate user evidence from an AI hypothesis
An AI summary can suggest what to investigate. It cannot turn a plausible response into an observed task outcome.
Read user evidence, critique the interface and explain what you would change.
Start a practice interviewAn AI summary can suggest what to investigate. It cannot turn a plausible response into an observed task outcome.
“AI is not raising the ceiling, humans are raising the ceiling.” (opens the source)
Polish is a decision, not a substitute for one. Name the task an AI-generated screen must support.
“break down our bigger vision into smaller, more testable pieces.” (opens the source)
Make a small prototype that tests a specific uncertainty before expanding the design.
Inspect labels, keyboard routes and status changes. A clean screenshot cannot show that a task works.
“But invisibility without trust is dangerous.” (opens the source)
A proposed location and fee need to be visible before booking. Keep a manual route when an AI suggestion fails.
Specify the component’s states and behavior. Test the fix where it is used, then say what remains unknown.
Monzo sources checked November 10, 2026.
Last reviewed Nov 10, 2026