# Before copying an enterprise AI pattern

- [ ] Open the primary source and preserve its date and attribution.
- [ ] Identify actual reported workflow, data and human role.
- [ ] Separate announced, planned, piloted and rolled-out states.
- [ ] Keep adoption, accuracy and business outcomes separate.
- [ ] Label vendor/company figures; do not invent comparable ROI.
- [ ] State what has not been independently verified.
- [ ] Choose a bounded local artifact and qualified reviewer.
- [ ] Review actual contract/settings, not a marketing assurance.
- [ ] Resolve access, connectors, training use, retention, deletion and logs.
- [ ] Use synthetic/approved material before production data.
- [ ] Define error/rework/escalation checks alongside time benefit.
- [ ] Record actual evidence before claiming success.

This checklist contains proposed checks, not completed test results.