From AI prototype to enterprise production
Before production approval, capture evidence for identity and authorization; code and data ownership; dependency review; integration; data residency and recovery; environments and rollback; observability and support; load and cost; and export of source, schemas, data, and operations.
The strongest exit test is practical: export or synchronize the application, deploy it through a separate pipeline where permitted, restore its data, rotate its secrets, and make a material change without the original authoring service. Code present in GitHub is valuable, but it is not by itself a complete exit architecture.