# AI operating cadence pack

Use this pack to keep production AI systems governed after launch. The core problem is not only release quality; it is whether product, platform, operations, risk, finance, and business owners keep looking at the same evidence and making explicit decisions.

## Cadence layers

- Weekly operating review: service health, support load, incidents, adoption, quality drift, cost, latency, and unresolved owner actions.
- Monthly value review: workflow coverage, cost per completed outcome, reviewer load, avoided effort, customer impact, and scale or hold decisions.
- Release gate: model, prompt, retrieval, policy, and tool changes with evaluation evidence and rollback paths.
- Quarterly steering review: portfolio value, funding, risk posture, and operating-model maturity across deployed AI services.

## Operating principle

Every review should produce one of six decisions: scale, tune, hold, rollback, fund, or retire. If the meeting only reports status, the cadence is not strong enough.