How we work
Align, Assess, Build, Sustain.
We build the capability in four steps. Each step leaves something your company keeps, and each one shows what the next step would need to close. You decide whether to take it.
What happens in each step
1Align
- What happens
- Management decides ownership, standards, risk and first steps.
- Offer
- Leadership Session
- What it leaves behind
- Decisions, owners, a 90-day plan
2Assess
- What happens
- A baseline of how your teams work with AI today.
- What it leaves behind
- Baseline, maturity per team, roadmap, business case
3Build
- What happens
- Engineers are trained, shared infrastructure is set up, and internal champions take ownership.
- What it leaves behind
- Trained engineers, team standard, skill library, champions
4Sustain
- What happens
- Adoption across teams, measured, with tools kept current.
- What it leaves behind
- Adoption in every team, measured evidence
You can stop after any step.
Each step is scoped and priced on its own. If you stop after one, you keep what it left behind: the decisions, the baseline, the trained engineers or the standard. Nothing is held back for a later phase.
How each engagement runs
Two ways to start
Most clients start with a Leadership Session or an Engineering Intensive. Trained engineers create the need for decisions; management decisions need people who know how. Whichever comes first, the other usually follows.A written next step
Every engagement ends with a written recommendation for what comes next, including what you can do without us.A baseline before a year
A new client starts with a Diagnostic before any 12-month commitment. The year is a management decision, made on evidence from your own teams.Handover by design
From the first month, internal champions take over more of the work. The aim is a capability that runs without us.
Terms we use
- Skill
- A written, reusable instruction that an AI coding agent can run, such as how to write a migration or review a pull request. Skills live in your repositories and are versioned like code.
- Subagent
- A separate AI agent that a main agent hands one bounded task to, with its own instructions and permissions.
- Orchestrator
- A skill that plans a larger piece of work and dispatches it to subagents, then checks their results before reporting back.
- MCP
- Model Context Protocol, an open standard for connecting AI agents to tools and data, such as your issue tracker or a read-only database.
- Hook
- A script that runs automatically before or after an agent acts, for example to block a forbidden command. Hooks enforce rules that instructions alone cannot.
Not sure where to start?
Tell us how your teams work today. We will suggest the first step, or tell you that you do not need one.