Is your AI product ready for engineering?

1–2. The user and decision are explicit

Name the person, their current moment, the decision or task that should improve, and the business consequence. “Add a copilot” is not enough.

3–4. The outcome and release boundary are explicit

Define the baseline, intended change, critical journey, and what is deliberately in, out, and later. The first release must be coherent without carrying the entire roadmap.

5–6. AI authority and failure are explicit

State what the model may propose or do, what requires human review, and what it must never do. Include recovery when the model, integration, or data fails.

7–8. Evaluation and ownership are explicit

Define representative tasks, acceptance evidence, telemetry, dependencies, decision rights, and the person authorized to accept or stop the release.