What it does
Runs the intake interview for a mobile feature request: asks one question at a time, scores each answer for whether it actually decided anything, and refuses to call the brief ready until the open decisions are closed. It exists because a department that produces a question LIST leaves the interview itself to whoever is calling it — four questions or sixteen, with nothing to say which was right. This asks the next question, one at a time, carrying the reason it is being asked and what changes depending on the answer, and it scores the reply by what the reply commits to rather than by matching it against a list of phrases. It then freezes the brief. A frozen brief carries the answers, the questions that were never answered, and every assumption standing in for one — so the work that follows is done against a boundary somebody agreed to rather than against gaps that were quietly filled in. It holds nothing between calls and says so. The brief travels as an argument, because a handle to storage that does not exist is worse than no handle at all. It never talks to the end user: it hands the caller a question to ask, and a caller who answers on the user's behalf has broken the only rule that makes the interview worth running.
Runs the intake interview one question at a time and freezes the brief.
Example prompts
- Start a brief: our onboarding loses 40% of users at the notification-permission screen.
- Here is the brief so far — what should I ask next?
- The answer was 'React Native, and only for the celebration modal'. Record it and score it.