You already know: how to scope and deploy into a customer's environment (previous lessons); evaluation pipelines and regression gates (D10); monitoring and on-call basics (D11).
Why this matters for Atlas. Most AI pilots that "succeed" never reach production. The demo worked, but nobody owned it, nobody watched its quality after launch, and the first silent regression ended the customer's trust. Production is a promise: it works on a bad day, someone notices when it degrades, and someone other than you can fix it. This lesson is how an FDE makes that promise real: a readiness review, a planned handoff, and support that catches quality drift before users do.
1. What changes between pilot and production
The question. The pilot passed. Why isn't it production yet?
Start from nothing. A pilot answers "can it work?" Production answers "will it keep working, for everyone, when nobody is watching?" Almost everything around the model changes: