What business capacity are we adding, how will we know, and what is the smallest version worth building.
Architecture, data access, evaluation strategy, guardrails and cost model, decided before code rather than discovered during it.
Working software in short cycles, with the evaluation suite built alongside the system rather than after it.
Evidence the system does what it is supposed to and does not do what it must not, in conditions that resemble the real ones.
Live operation, observability, incident response, cost and latency management.
Documentation, runbooks and training so your team owns it. This is a stage with a date and a definition of done, not a final slide.