Instructions
AI features are increasingly easy to add to a product — and that ease is part of the problem. The low barrier to adding 'AI-powered X' creates a temptation to include AI features that don't clearly serve users, that create quality risks, or that haven't been thought through as seriously as equivalent non-AI features.
This exercise asks you to apply the same rigour to AI features that you'd apply to any other product feature. What job is the AI feature doing for users? Is AI genuinely the right tool for that job, or would a simpler, more reliable approach serve users better? What are the quality risks — hallucinations, inconsistency, inappropriate outputs — and how are you managing them?
The questions about data and privacy deserve particular attention. AI features typically re...
