What I am building.

I lead enterprise AI transformation, and I build too. I create working tools, workflows, analytics, and open source to test ideas and make deployment repeatable.

What I am working on.

AI coding adoption playbook

Public

A practical rollout kit for moving AI coding tools from access to regular use across engineering teams. It has guides, templates and a small adoption scorecard.

Agent rules and skills

Partly public

The rules, reusable skills and review points that tell Cursor and Claude Code how to work in my repos. They live in the repo, not in chat history. Most are private. A first set is public as Cursor plugins. They are early and not yet submitted to a marketplace.

This website

Public

The code is MIT-licensed. The site is prerendered for static hosting. The build fails if a route or an asset is missing, so broken pages never ship.

Content pipeline

Private

Turns a brief into a published post through fixed steps. It has a quality check, and nothing goes live without my approval.

Keynote OS

Private

Stores the talks I have delivered and pulls out reusable stories, frameworks and data points for the next one.

AI Ecosystem

Private

A 12-month system for learning the AI stack. It holds dossiers on the tools and companies I follow.

Built with Cursor and Claude Code, with human review, tests and guardrails.

What the scorecard prints.

A small tool in the playbook repo. It reads a per-user usage export and prints each signal on its own. There is no composite score and no outcome claim. This run uses a simulated 12-user pilot.

Adoption scorecard, synthetic six-week pilotSynthetic data, week 6 of a simulated pilot with 12 users. Funnel: 12 licensed, 11 activated, 7 weekly active, 6 sustained. By team, sustained users: team-api 5 of 7, trending up; team-tools 1 of 5, trending down. No composite score. No outcome data in this export. All numbers are invented.ADOPTION SCORECARDSYNTHETIC DATA · WEEK 6 · 12 USERSLicensed12Activated11Weekly active7Sustained6SUSTAINED, BY TEAMteam-api5 of 7, upteam-tools1 of 5, downNo composite score. No outcome data here.Invented numbers. No real organization.
The same week 6 data, drawn as a funnel and a team view. Synthetic data. Invented numbers.
$ python -m tools.adoption_scorecard \
    data/worked_example_week6.csv \
    --previous data/worked_example_week3.csv

# Adoption scorecard: worked_example_week6.csv

As of: 2025-03-14
Users read: 12. Rows skipped: 0.

## 1. Funnel

| Stage | Users | Share of licensed |
| --- | --- | --- |
| Licensed | 12 | 100% |
| Activated (any use) | 11 | 92% |
| Weekly active (last 7 days) | 7 | 58% |
| Sustained | 6 | 50% |

## 2. By team

| Team | Users | Sustained | Sustained rate | Median active days | Trend |
| --- | --- | --- | --- | --- | --- |
| team-api | 7 | 5 | 71% | 12 | up |
| team-tools | 5 | 1 | 20% | 3 | down |
Real output of the open-source scorecard on the simulated worked example in the playbook repo, shortened to the funnel and team table. The full report adds depth, a status panel and what each metric can hide. Invented numbers. No real organization.

Results from enterprise rollouts.

2.7×
Growth in weekly active GitHub Copilot users in one R&D unit over three months
5×
Scale-up of an AI-enabled solution in 8 weeks
31
Estimated engineer-years of annual capacity from 560 hours of effort

Want to talk about any of this? Message me on LinkedIn or by email.