Problem Fit
Ready:
AI is justified by the workflow, not by novelty or model availability.
Not ready:
The system starts from a demo or capability claim.
Release Scorecard
Five questions before production launch.
A concise readiness gate for deciding whether an AI system should launch, pilot, narrow scope, or stop.
Boundaries
Ready:
Inputs, outputs, data access, tool authority, and human approvals are explicit.
Not ready:
The prompt or model is expected to infer product policy.
Grounding
Ready:
Sources are authorized, fresh enough, trusted, and handled honestly when missing.
Not ready:
The model is pressured to answer without required evidence.
Evaluation
Ready:
Release evidence matches the system's risk, autonomy, and user exposure.
Not ready:
Confidence comes from demos, anecdotes, or generic model benchmarks.
Operations
Ready:
The system can be traced, rolled back, monitored, costed, and improved after launch.
Not ready:
Infrastructure health is treated as proof of product correctness.
Decision outcomes
Do not treat readiness as only yes or no.
- Proceed when all five areas have evidence and owners.
- Pilot when uncertainty remains but blast radius is bounded.
- Narrow when the decision surface or autonomy level is too broad.
- Block when consequence is high and evidence, control, or accountability is weak.