AI-Accelerated Engineering Leadership Checklist (Factory Mode)

Purpose
Turn AI-assisted output into reliable production results. This checklist is designed to be enforceable: each item should map to a CI check, an access control, a required artifact, or a named owner.

1) Branch and CI gates (non-negotiable)
- Main branch protection is enabled (required checks must pass).
- Tests run automatically on every PR.
- Linting/formatting runs automatically (no “style” debates in review).
- Type checking or static analysis is in place where the language supports it.
- Secret scanning runs on every PR (and ideally on push).
- Dependency scanning is enabled (Dependabot/Snyk or equivalent).

2) Human review focuses on what automation can’t catch
- PR template forces: intent, risk, rollout plan, and customer impact.
- Reviewers are instructed to ignore generated code volume and focus on:
 - Correctness of logic and edge cases
 - Data handling and privacy implications
 - Security-sensitive flows (auth, permissions, injection surfaces)
 - API/interface stability and backward compatibility
- Set a norm for change size: prefer multiple small PRs over one large PR.

3) AI usage rules that can be enforced
- Approved AI tools/accounts are defined (enterprise accounts where possible).
- Data classification rules exist: what is allowed vs prohibited to paste into external tools.
- Logging/audit expectations are defined for approved tools where available.
- A fallback path exists: how work proceeds if AI tools are unavailable.

4) Release discipline
- Feature flags or staged rollouts are standard for risky changes.
- Rollback is documented and tested (not theoretical).
- A launch checklist exists and is required for production changes.
- Observability is present: logs, metrics, traces as appropriate for the service.

5) Ownership and accountability
- Every service has a named owner (team or individual) with clear escalation.
- On-call responsibilities are real (rotations, paging, runbooks).
- Post-incident reviews produce concrete changes: tests, monitors, guardrails.

6) Two weekly leadership reviews (lightweight, high signal)
- Decision review: resolve one real tradeoff with a decision owner.
- Quality review: look at regressions/incidents and identify one gate to strengthen.

How to use this
Pick one critical service. Apply sections 1–6 end-to-end. Do not expand scope until the first service is “boring” in production: predictable deploys, clear ownership, and fast rollback. Repeat service by service.