Agentic engineering, installed on your team

Go from "we tried an AI agent once" to a disciplined agent team that ships production code: safely, test-first, and reviewably.

Your developers already use Claude, Cursor, or Copilot. Some ship faster, but reviews are inconsistent, an agent refactored half a service unsupervised, and you can't tell your board whether it's a productivity win or a liability. I install the discipline that turns that into safe velocity: a five-role agent team, test-first, with gates a human signs off. Every course runs on a shared sandbox codebase, so a room of people from different companies all work on the same real thing.

The catalogue

Two lines, six ways in

Agentic engineering

1 day, open enrolment

One seat · classroom · weekday or Saturday

One developer, one seat, on a shared sandbox codebase the whole room works from. You run the orchestrator-first loop, tighten tools to least privilege, TDD a real slice behind a QA gate, and take a working agent team home.

You leave with the practice on your own machine, not notes about it.

1 day, private

Your team · classroom or your offices

The same day, run for one team, on the sandbox by default or your own repo on request. The room is yours, so the examples and the questions are yours too.

One team, one shared practice, agreed in a day.

2-day intensive, private

Your team · the full practice

Everything in the 1-day, plus safe tools (MCP), parallelism with isolated worktrees, and your team's written "when we use agents, and when we don't" policy with a human-oversight checklist.

A teamwide, auditable practice, not one enthusiast's trick.

EU AI Act and compliant AI

Executive briefing

90 minutes · leadership · no code

What the EU AI Act actually asks of you, what it does not, and which of the two your organisation is exposed to. Plain language, current dates, no scare slides.

Your leadership can tell an obligation from a rumour.

EU AI Act, full day

Your team · non-technical room

The working session: scope your systems, the duties that apply now, the human-oversight evidence an auditor accepts, and a 90-day plan you leave with.

A written plan with owners, not a folder nobody opens.

EU AI Act, 2 days, regulated sectors

Regulated sectors · wider scope

The full day plus the high-risk deep dive: classification, documentation, the oversight and record-keeping duties, and how they land on the systems you already run.

An evidence trail mapped to your own systems.

Something else

Run it on our own codebase

The classroom teaches on a shared sandbox. If you want the same work done on your own repository, say so and I'll scope it.

Embedded advisory or pilot

A few weeks part-time on a real project, installing the practice for real, until the team is self-sufficient.

Why me

A ~26-year production architect who built a multi-agent SDLC as a real platform, teaching it on your repo, naming the limits, showing the audit trail. Not a course-first creator, not an influencer. The calm veteran a CTO trusts with their team.

Bonus: the gates, ADRs, and audit trail are the same shape of human-oversight evidence the EU AI Act asks of high-risk systems, and the kind of thing customers, auditors and procurement increasingly ask of everyone. Compliance falls out of good engineering, not a separate workstream.

Want one of these scheduled?

Pick everything you're interested in and leave your email. When enough people want the same thing, I schedule it and you hear first.

What are you interested in?

Agentic engineering
EU AI Act and compliant AI
Something else

Only used to tell you when a session is scheduled. No newsletter, nothing else.

Book a scoped call

Every engagement is a fixed-scope package, quoted to your team and goals. Book a 20-minute call and I'll scope it.

Book a 20-minute call