Marty Cagan's framework was written for human product teams, but its core demand — that someone is genuinely accountable for outcomes, not just delivery — hits harder when your delivery mechanism never sleeps and never asks whether this was a good idea.
Cagan's argument in Inspired is not about process ceremonies or roadmap formats. It is about the difference between a team that takes a problem and figures out the right thing to build versus a team that executes a list someone else handed them. The four tests — is the solution valuable to the customer, usable by them, technically feasible, and viable for the business — are meant to run continuously, in discovery, before a single line of production code gets written. The failure mode he diagnoses is the feature factory: output mistaken for progress.
When agents handle the execution layer, the feature factory trap gets worse, not better, because the cost of shipping something wrong drops just enough to feel safe. Cagan's four tests do not change, but ownership of them has to be re-assigned deliberately. Your agent can run feasibility checks and instrument usage; it cannot tell you whether this solves something a customer actually cares about. That judgment lives with a human, and in an AI-native studio it has to be structurally protected — not assumed to happen in the margins between deploys.
- Continuous discovery is not a phase, it is a standing obligation your agents accelerate delivery without replacing
- the four tests are your founder's checklist before any agent loop gets authorized to ship
- outcome ownership must be named explicitly or agents will optimize for the metric you measured, not the problem you meant to solve.
