Production-ready AI (production readiness)
An AI feature that is tested, monitored, capped in cost, limited in permissions and can be rolled back, not one that only worked in the demo.
A production readiness review is a checklist before launch. For AI it usually covers evals on real cases that run on every change, LLM observability per request, a cost cap, a pinned model version, permissions that match the user, a fallback when the model fails and an owner on call. Most of the distance between a pilot and production is this list, not the model.
Related terms
- Pilot purgatory (PoC purgatory): When AI proofs of concept and pilots keep running or get dropped without reaching production, or go live but cost more in human oversight than they save.
- Evals: A repeatable test set that scores an AI feature, so changes can be measured instead of guessed.
- LLM observability: Tracing every model call with inputs, outputs, latency, tokens, cost and the user outcome.
- Cost cap: A hard spending limit enforced by your code, per user, per day or per month.
- Model pinning (model versions and deprecation): Calling an exact, dated model version instead of an alias like “latest”, so behavior does not change under you, and planning for its retirement date.
- Guardrails: Checks around a model that block, fix or flag inputs and outputs that break the rules.
