The AI Product Lifecycle

Traditional product discovery asks 'should we build it' — and feasibility was roughly a constant, because engineers could estimate. AI features invert this: feasibility is the unknown variable. The question isn't 'can we build it' (anyone can wire an API) but 'can a model do this well enough, at acceptable cost and latency, for this use case'.

Discovery now starts with capability probes, not just user interviews — you interview to find the problem, then probe the model to learn whether the problem is solvable this quarter.