AI tooling leadership across the engineering organisation

Career·Updated

AI tooling leadership across the engineering organisation

A cross-organisation operating system for AI enablement that Matt built and maintains. In 2026 it matured from evangelism into governed distribution infrastructure (the plugin marketplaces) and autonomous agents. This document covers the community layer.

Talks

Internal talks date back to a 2019 talk on the backend-for-frontend and GraphQL pattern. The 2026 engineering brown-bag series includes three of Matt's:

  • February 2026, "Agentic Engineering in 2026," originally titled "Coding Agents: Is this the end of writing code by hand?"
  • May 2026, "Effective Context Management for AI Agents." Thesis: frontier context windows are now around a million tokens, but the engineering skill that matters is what you put in the window, not how big the window is. A 42-minute talk with six interactive HTML teaching artifacts.
  • August 2026, "Advanced Context Engineering: Give your agent a brain," on building plugins that give agents a knowledge base persisting across sessions.

Earlier: a July 2025 lunch-and-learn on practical use of Cursor, an April 2026 demo to the company's AI guild, and a quarterly AI tech-talk cadence agreed with leadership.

Recurring demos and forums

  • Weekly developer-experience demo series: a recurring presenter through summer 2026, demonstrating the security-fix skill, adversarial review, the subject-matter-expert plugin scaffold, the team marketplace, and the autonomous remediation harness and its learnings. One service's independent-deployment epic was generated live during a July demo. His framing at the last of them: any of the experiments can be adapted to many other use cases.
  • The organisation's AI-tools forum: founded by Matt in February 2026 as the cross-organisation channel for questions, access requests, troubleshooting, and knowledge sharing. It is where the marketplace alpha was announced and where the marketplace governance feedback landed.
  • A short-lived newsletter: three monthly editions covering models, tools, and industry articles, retired in early 2026 in favour of sharing updates as they happen. Lower ceremony, higher frequency, more conversation.

The skills library

Matt created and maintains the organisation's agent-skills library, the distribution mechanism before the marketplaces. The skills for drafting tickets and epics were the breakout hit; one engineer reported that they turned three to four hours of work into a thirty-minute conversation, and an early update to organisation leaders credited them with saving teams significant time.

Access, budget, and proposals

  • Owns the organisation's AI tooling access and spend for agent workflows.
  • Authored proposals to leadership: an AI-champions-per-team adoption proposal; the plugin marketplace governance proposal; an LLM evaluation-framework proposal (February 2026, local eval tooling first with managed evaluation later); and testing-workflow proof-of-concept notes written with peers elsewhere in the organisation.
  • Invited onto organisation-wide agentic-engineering interview panels.

Evaluating tools honestly

  • Led a trial of a third-party coding agent in late 2025 that concluded against purchase. The trial ran as a structured internal programme with nine developers, trial questions, two weekly check-ins, and a survey.
  • Led the Claude Code trial whose results informed the organisation-wide decision to adopt Claude Code and Cursor as primary tooling for roughly fifty engineers, and traded in his own seat on the other editor to pay for it.
  • Enabled newer models in the company's existing AI account, coordinated network allowlisting, and pulled usage analytics for cross-organisation adoption measurement with three peers.

On his own team

The Email Reliability team has 100 percent adoption of Claude Code for code generation, with AI review on every pull request; every engineer on the team has access, and Matt works alongside them using it daily. The commit-trailer telemetry is a floor on that, not a contradiction: it counts one tool's signature and the other editor leaves none.

Where he wants to push next

Named openly as unfinished rather than claimed: making the remediation agent a hosted pattern any team can run; broadening distribution and publishing adoption evidence; exporting the proof-of-concept-to-production playbook and the eval-then-upgrade discipline to at least one other team; and taking the two context talks outside the company to a meetup and a conference call for papers.