# Regulated industry rollout map

Use this map when an AI program touches regulated, public-service, safety-critical, or operationally sensitive workflows. The goal is to adapt the rollout to the domain before architecture, procurement, or change-management decisions harden.

## 1. Domain constraints

- Name the industry, workflow, accountable service owner, and decision classes that must remain human-owned.
- Record policy, regulatory, accessibility, safety, data-residency, and customer-impact constraints.
- Identify review participants across operations, technology, risk, legal, security, and frontline teams.

## 2. Evidence model

- Define the source systems, policy references, records, user actions, and reviewer decisions needed for each workflow class.
- Separate evidence used for recommendations from evidence required for approvals.
- Keep generated summaries traceable to source material, timestamps, and responsible owners.

## 3. Release gates

- Start with bounded cohorts and explicit rollback criteria.
- Compare assisted, automated, and manual paths by quality, cycle time, escalation load, and user confidence.
- Treat unresolved control gaps as launch blockers, not post-launch cleanup.

## 4. Operating review

- Review adoption, exceptions, quality, cost, incidents, and stakeholder feedback on a fixed cadence.
- Decide where to expand, tune, hold, or retire workflows using the same evidence model used during release.
- Keep audit artifacts current so the system remains explainable after the initial pilot.