Checklist
Production AI Checklist
A practical checklist for enterprise teams moving from GenAI pilot to production: data, security, evaluation, ops, and ownership.
Use this in workshops, security reviews, and delivery planning. Print or copy freely.
Problem and outcome
- ✓Business outcome is measurable and owned by a named stakeholder
- ✓Success criteria are defined beyond “users liked the demo”
- ✓Out-of-scope uses are explicit
Data and permissions
- ✓Data sources and sensitivity classes are inventoried
- ✓Access control model matches source systems
- ✓Retention, logging, and residency requirements are documented
Architecture
- ✓Vendor-neutral target architecture is written down
- ✓Tool and API boundaries use least privilege
- ✓Failure modes and fallback behavior are designed
Evaluation and release
- ✓Golden dataset or evaluation set exists
- ✓Quality, safety, latency, and cost gates are defined
- ✓Regression checks run before each release
Operations and handoff
- ✓Monitoring and alerting cover critical failures
- ✓Runbooks exist for incidents and model/tool changes
- ✓Internal owners can operate the system without the delivery team