AI feature sensemaking
Judge whether an AI idea solves a real workflow problem or just adds novelty.
The demo always looks magic. The question is whether anyone still opens it in week three.
A stakeholder wants an "AI assistant" in your project-management tool that can "answer any question about your projects." The prototype demos beautifully and leadership is excited to ship it.
What's the strongest read?
Capability is not a job
AI makes it cheap to build things that look impressive in a two-minute demo. That's exactly the trap. Impressiveness is a property of the capability; usefulness is a property of the workflow it sits in. The two come apart constantly — which is why so many AI features spike on launch and flatline by the second week.
The discipline is the same one that governs every feature: start from a user with a job to do and friction in the way. AI is a means. If you can't name the specific moment it removes friction, you have a capability looking for a problem — and users will feel that gap even when the demo doesn't show it.
Novelty framing vs workflow framing
Novelty framing
Starts from what the model can do.
- "It can answer anything."
- Success = an impressive demo
- Scope is open-ended and vague
- Retention quietly craters after launch
Workflow framing
Starts from where the user is stuck.
- "Users lose 10 min hunting for status."
- Success = that task gets faster
- Scope is bounded by the job
- Usage holds because the friction was real
Don't ask what the AI can do. Ask which specific, recurring moment of friction it removes — and whether users would miss it if it disappeared. If you can't answer, it's novelty.