Function-by-function AI rollout sequence — engineers first, non-eng last

Observational sequence from one mid-stage rollout: engineers self-migrate within 1-2 months once tooling is available; engineers then drive QA migration; SRE resists longest but eventually adopts under pressure from concrete results; second-line (IT, CRM, BI, helpdesk) and marketing lag because git is the entry barrier for them. Initial resistance + minor sabotage exists at the very start, then dissipates.

This is a planning artifact, not a prescription. Knowing the order lets you sequence training spend, time hiring filters, and predict which function will scream when. The reported timeline: full org migration in months, not quarters — but only if you don't reverse course when SRE pushes back.

1. Start with engineering. They adopt fastest because their existing tooling (Cursor, Claude Code, Codex) is already AI-native.
2. Let engineers champion QA. Once eng is productive, QA migrates next under peer pressure rather than top-down mandate.
3. Apply pressure with concrete results to SRE. They will resist longest. Don't fight philosophically; show the numbers from eng + QA.
4. For non-eng functions, lower the git/CLI barrier first. Pick tools that don't require git literacy (Cowork-style hosted surfaces, MCP-wrapped data sources). Without this, marketing/legal/CS stalls indefinitely.
5. Expect early resistance and minor sabotage. Don't reverse course on the first wave of pushback — it fades within a few weeks.

• Git is the wall for non-engineers — not "AI skepticism." Multiple operators initially diagnosed non-eng adoption failures as resistance; the actual blocker was that branching, commits, and PRs are foreign. Pick tools that hide git from marketing, CS, and legal entirely, or budget weeks for git literacy training before expecting AI adoption.