Let an AI Agent Map Interview Evidence—Not Make the Product Decision

Idea: Give early product teams an agent that turns customer-interview notes into evidence-linked themes and open questions—not a list of features to build.

The agent could group repeated problems, but every theme must point back to the specific notes that support it. A product lead reviews the evidence, corrects the grouping, and decides what—if anything—changes in the roadmap.

Start with a small, consented batch of notes the team is already allowed to use. The riskiest assumption is that the synthesis saves review effort without blurring distinct customer needs. Track whether reviewers can verify each theme and how much correction it takes. Continue only if the evidence remains traceable and the team finds the review useful; stop if summaries hide disagreement or require more repair than reading the notes.

The agent can open a path through messy feedback. It shouldn’t choose the destination.