Most AI adoption plans hand everyone the same ladder. But a CIO adopts AI through funding and risk appetite, a business analyst through drafted stories, and an on-call engineer during an incident at three in the morning. This is a working version of a strategy that treats them differently: an adoption ladder per role, a risk tier per use case, nineteen role tracks, and one path from requirement to production. It is written for the hardest case, a regulated organization, and it is a shape, not a final standard: take it, then tune it to your own workflow, controls and team size.
Each role shows five steps: two first steps, two that grow the practice, and what fully adopted looks like. Then it shows what changes for the organization. Pause on any role.
Many organizations buy a tool, hand out seats and wait for value. Usage rises, then flattens. Pilots look good but never reach production, because risk sees them for the first time at the end. This strategy puts the order the other way round: governance first, shared team context second, delegation to agents last, where the controls are proven.
An adoption ladder per role (Stage 0 Foundations to Stage 5 Scale and sustain) and a risk tier per use case (low, medium, high, prohibited). A high stage is never permission to automate a high-risk use case.
From executive sponsor to production support. Each track opens with first steps, then the fully adopted workflow, what changes for the organization, and how a small or unregulated team covers the role.
Every role on one path: ask, shape, build, test, release, run. At the start, AI only drafts. Fully adopted, agents draft, check and assemble evidence, and a named person still makes each accountable decision.
Agents check the asset catalog before writing a helper, and CI enforces what local hooks only advise. One worked example runs through every page, with labs, templates and a glossary behind every term.