This is a knowledge site on adopting AI stage by stage, role by role. It is written for the hardest case, regulated teams, and it applies to any organization that wants AI to reach production safely. Treat it as a working version: a shape to tune to your own workflow, not a fixed standard. It is general guidance for engineering leaders, AI-enablement leads and team leads who must make AI useful without bypassing the controls they already answer to. It is built from practice: the patterns come from real delivery work, written up as one fictional organization and one worked example (the contractual payment due date and the payment cutoff).
What it is and is not
| It is | It is not |
|---|---|
| Patterns for governance, adoption, reuse and measurement | Legal, regulatory or compliance advice |
| Templates, each shown blank and filled | A certified control framework |
| Labs that run in any approved coding agent | A product pitch for one tool |
| One worked example, followed from Stage 0 to Stage 4 | A record of any real organization |
Who wrote it
Written by Sam Muthu, engineer and architect.
Sibling sites
| Site | What it covers |
|---|---|
| ask-to-release.sammuthu.com | What delivery looks like at the top of the ladder: one ask, front-loaded artifacts, release |
| claude.sammuthu.com | An architect's study guide to building with Claude |
| scaling.sammuthu.com | Scaling agents: patterns for running many agents well |
How it was built
| Step | What happened |
|---|---|
| Static site | Plain HTML generated from markdown and JSON; no accounts, no sign-in |
| Locked spec | The structure, example world and terms were fixed before any page was written |
| Multi-model review | Every page was reviewed by multiple AI models for accuracy and gaps |
| Human review | A human read and approved every page before publishing |
Disclaimer
This site offers patterns, not legal or regulatory advice. Scope and terms, such as what counts as a "model", follow each organization's own policy and jurisdiction. The organization, teams, services and examples are fictional. Counts, durations and percentages are illustrative, not measured results.