What changes when you take an AI prototype to production?

Almost everything except the core idea. Production adds real integrations and authentication, security and compliance reviews, robust handling of edge cases and failures, evaluation and regression testing, monitoring and alerting, scalability and cost controls, human-in-the-loop workflows, audit logs, documentation and runbooks, user training, and ownership and support processes. The prototype proves value; production makes it reliable, safe and maintainable.