Mental model
AI product discovery starts from a costly decision or workflow, then tests whether AI capability fits the task under real constraints.
Learning outcomes
- Explain the mechanism and ownership boundaries behind AI Product Discovery.
- Compare the main design alternatives and their operational trade-offs.
- Diagnose common failures and select evidence for a production decision.
Theory
Map actors, decisions, current failure cost, available context, reversibility, feedback, risk, and adoption friction. Test the narrowest valuable capability before designing a general assistant.
Trade-offs
Narrow workflows produce clearer evidence and controls but may appear less ambitious. Broad assistants promise flexibility while hiding evaluation and ownership boundaries.
Failure modes and misconceptions
Starting from a model demo; no baseline; measuring engagement instead of outcomes; ignoring correction work; selecting irreversible tasks; and treating user excitement as capability evidence.
Decision scenario
A team proposes an AI assistant for all finance work. Narrow the idea into a testable workflow with measurable value and acceptable risk.
What evidence distinguishes an impressive model demonstration from a viable product workflow?
Primary sources
human-ai-guidelinesnist-ai-rmf