Model Deployment Readiness Checklist
Check before a model goes live.
The Open buttons send the prompt text as currently shown below to that site only when you click. Fill in the placeholders first.
[VERIFY] is an editorial flag, not a blank to fill. Wherever the prompt shows [VERIFY], check that item against a reliable source before you use it.
Create a deployment readiness checklist. Model and use: [MODEL_USE]. Environment: [ENVIRONMENT]. Risk level: [RISK]. Cover data (training data provenance, leakage checks), performance (metrics on holdout and slices), robustness (edge cases, input validation), reproducibility (versions, seeds, code and data snapshots), documentation (model card), monitoring, rollback, human oversight, privacy and security review, and ownership. For each give the check, evidence required, and owner. Add go/no-go criteria with thresholds I set. Include a staged rollout plan with shadow mode and a first-week review. Mark regulatory points [VERIFY]. Add a one-page sign-off sheet with roles for the data scientist, engineer, product owner and risk reviewer, and a section for known limitations that users must be told about. Include a post-launch review at one week and one month with the metrics to check. Add a worked example of two checklist rows written in full for an invented churn model: data (check: training data snapshot recorded with hash; evidence: manifest file; owner: data engineer; threshold: complete) and performance on slices (check: recall within five points of overall for each region; evidence: evaluation report section 4; owner: data scientist; threshold set by product owner). Label it an illustration. Add a staged rollout table with percentages, duration, metrics watched and the rollback trigger for each stage, and a first-week review agenda. Provide a model card outline of eight headings (purpose, data, performance, limits, fairness checks, monitoring, contacts, version history) and a rule that no model goes live without a named owner on call. Also include a list of the artefacts to archive at release, such as the model file, the training data snapshot, the code version and the evaluation report, and who holds them. Add a short rollback rehearsal: how to switch back to the previous version and how long it should take. Finish with a note on how to inform users about what the model can and cannot do.
Fill in before use
Optional: type your details and the prompt above updates. Nothing is saved or sent.
Replace each bracketed placeholder with your own details, then paste into Any AI or another assistant.