Skip to content

OnDynamics Labs

Where I test new models and techniques before they reach a client programme.

Contact

AI Enablement for IT Teams

GitHub Copilot and Claude Code, for your whole IT team.

I help your developers, testers and IT engineers use GitHub Copilot and Claude Code on their own backlog, with permissions your security team signs off and a review and tests on every change.

  • GitHub Copilot
  • Claude Code
  • MCP
  • Playwright
  • Azure DevOps
What I do for you

Coding agents, with the controls your CISO asks for.

With good automated tests and small pull requests, coding agents speed a team up. Without them, the extra code arrives faster than anyone can test it. So the tests, reviews and permissions come first:

  • Guardrails first. Enterprise policies in GitHub Copilot, managed settings in Claude Code, approved MCP servers, secret scanning and an audit trail, agreed with your security team before anyone starts.
  • Small, reviewed changes. Shared project instructions, a plan before every task, small pull requests and tests on every change. A person reads every change before it merges.
  • Your own backlog. Your team applies it to its own backlog: legacy analysis, integrations, test automation with Playwright, upgrades and migrations.
  • Measured. Adoption, lead time, change failure rate and review load, before and after, so you can see what changed.
Why it matters

Nearly everyone uses AI. Almost a third don’t trust its code.

90%

of nearly 5,000 technology professionals use AI at work1

30%

have little or no trust in the code it generates1

NCSC

Understand the risk, then set how much freedom the AI gets: least for authentication, secrets and sensitive personal data2

  1. DORA, State of AI-assisted Software Development, Google Cloud, September 2025.
  2. National Cyber Security Centre, The ‘vibe coding spectrum’ approach to AI-assisted software development, 18 June 2026.
Where I’ve done this

GitHub Copilot and Claude Code for a 15-person team of Microsoft and partner engineers I guided at Microsoft, AI-run Playwright testing at a UK financial services organisation, and this site and Arc, its AI assistant: built with Claude Code, powered by the Claude API and other LLMs.

  • GitHub Copilot
  • Claude Code
  • Claude API
  • Playwright
  • MCP
  • Threat modelling
Ways to start

Four ways to start.

AI coding assessment

How your team builds and ships today, where agents will help first, and the guardrails to agree before they start. 2 weeks, at a fixed scope.

Ask about this

Guardrails setup

GitHub Copilot and Claude Code set up for your organisation: policies, approved MCP servers, managed settings, audit and budgets. 3 to 4 weeks.

Ask about this

Team enablement

Your team learns on its own backlog with me alongside: pairing, reviews and a weekly clinic. 8 to 12 weeks.

Ask about this

The six-day course

Legacy to live with AI agents, for up to 12 people, on a fictional bank’s CRM 4.0 system.

See the course
Bringing coding agents into your delivery?

Start with an AI coding assessment.