Where does a PM's prototype stop?

Product manager Uditi Sharma argues that AI prototypes show how a feature feels but do not replace the written document that explains scope and success.

Uditi Sharma writes a newsletter about product management for AI products. When preparing for a senior product manager interview at Hopper, a travel booking app, she built a working AI booking assistant and wrote several versions of the requirements document. She concludes that each did a job the other could not.

• AI features need prototypes more than ordinary features, because the AI model's answers, delays and failures are the product and a static mock-up cannot show them.
• She suggests testing the AI model's real instructions on about 20 real inputs in a chat tool before building any interface.
• Each prototype should start from a written card naming its one question, the result that would change the plan, who tests it and a time limit.
• She leaves open who owns the AI model's instructions after the prototype passes to engineers, and whether prototype code should reach production.

Product teams can ask that each prototype names the decision it is meant to change before anyone builds it. The piece describes one product manager's practice and interview preparation, not a team's measured results.

Derived by Working Surface from the article; more in the Playbook. Source line: Demos are fine for selling an idea. They are poor at changing a decision.

15 October 2026: when AI agents make building cheap, the decision worth reviewing is the one made before the build, written where other people can dispute it.