Why does an AI-generated prototype need a sign-off list at all?

Because it contains choices nobody made. Ravi Mehta found features he had not chosen, assumptions he had not made, and decisions he had not reviewed in an AI-written PRD, and prototypes carry the same thing in copy, defaults, empty states, and uncertainty. Those choices look finished, so nobody argues with them, and they ship by default.

Does the PM own a prototype the PM built?

No. The PM owns the question spec, the problem, the outcome, and the one question the prototype answers. The prototype is evidence for that spec and for design's behavior spec, and it belongs to neither. Who built it does not change who decides.

Who owns the system prompt after handoff?

Split it by line. The PM owns the lines that say what the feature is for. Design owns the lines that set behavior: what it must never do, its tone, and what it says when it is unsure or refuses. Engineering owns implementation, versioning, and tests.

How long does the ledger take?

About thirty minutes for a typical prototype. List every decision, not every feature, then sort each into answered the question, behavior, or made by the model. The first pass usually shows a third column longer than anyone expected, and most of it is behavior.

How does this relate to the design rubric?

The rubric is the gate, and the ledger feeds it. A rubric scores what is on the page. The ledger makes sure the page includes the decisions nobody noticed making, so the rubric has something to score.

Who Signs Off on Decisions in an AI Prototype?

Who signs off on the decisions inside an AI-generated prototype?

THE SHORT ANSWER

Sort the prototype's decisions into three columns and give each column one owner. Choices that answer the prototype's question are the PM's to ratify or kill. Behavior choices, meaning what the surface does, says, refuses, and does when it is wrong, are design's to ratify or turn into a spec line. Everything the model chose that nobody chose stays in a third column and does not ship until an owner moves it. The prototype itself belongs to neither function, whoever built it.

The PM can generate a working prototype in an afternoon now. That doesn't settle who decides what's in it.

My rule, from part 2 of The Design Transition, Your Prototype Made Decisions Nobody Made. Who Owns the Spec Now?: after every prototype the team reacts to, list its decisions and sort them into three columns.

Answered the question. Problem, user, outcome. The PM ratifies or kills.
Behavior. What the surface does, says, refuses, and does when it's wrong. Design ratifies or writes the spec line.
Made by the model. Everything nobody chose. It doesn't ship until an owner moves it.

The boundary for the PM comes from Uditi Sharma: the PM's prototype stops when it has answered its one question. Everything past that is behavior, and behavior is design's to specify.

This refines a rule I set in May, in Kill the PRD. The Prototype Is the Spec.: a prototype in four to six hours, a one-page rubric with three to five criteria and at least twenty test cases. The prototype is the spec only for the choices someone made on purpose. The ledger catches the rest.

Every part of the series is at the series hub, and the gate the ledger feeds is in Should designers be the quality gate for PM prototypes?

Last reviewed 2026-10-21 · 2 min read

Who signs off on the decisions inside an AI-generated prototype?

SOURCES

THE LONG VERSION

RELATED ANSWERS