AI enablement · ActiveCampaign
Turning AI adoption into measurable, repeatable practice
At ActiveCampaign, I led AI enablement and transformation across engineering, product, and leadership. I paired adoption goals and operating metrics with accessibility-first learning experiences that made practical experimentation possible.
- My role
- Senior Program Owner · AI enablement and transformation.
- Context
- ActiveCampaign, June 2025–September 2026. Product-manager roadshow in March 2026.
- My deliverables
- Devin goals and KPIs, hands-on Cursor learning, Wavelength, Claude skills, and evaluation prompts.
Evidence
Measured transformation
- Adoption
- Agentic-coding adoption moved from roughly 2% to 43% of merge requests across the broader transformation work.
- Enablement surface
- Led a hands-on product-manager roadshow across three continents.
- Behavior change
- Product managers began producing prototypes of their own after the roadshow.
The adoption metric reflects broader organizational work. I contributed to that result; I do not claim the training program alone caused the increase.
Engineering & product
What the artifact taught me
What I was actually building
The product was not a workshop. It was an operating system for adoption: goals and KPIs, practical learning experiences, visible artifacts, decision structures, and enough psychological safety for people to try the tools on real work.
The repeated pattern was to pair permission to experiment with a way to inspect the result. That is how a novelty starts becoming a practice.
The problem
ActiveCampaign had a strong push toward AI adoption, but little shared structure for how engineering, product, and leadership would learn, measure progress, and act on it.
What I owned
As Senior Program Owner, I led the Devin adoption effort: goals, KPI measures, and a game that explained the operating model. I also owned the hands-on product-manager roadshow and built tools for shared program state and AI-response evaluation.
What I did
I paired the Devin adoption target with explicit goals and KPI measures, then coordinated cross-pillar work around outcomes and north-star metrics for executive review.
In March 2026, I led a roadshow for product managers across three continents, introducing Cursor and agentic AI. Product managers began making prototypes afterward.
My Cursor boot camp gave product people room to be beginners: experiment, build something, and learn from what happened. Participants requested GitLab access so they could keep going.
I built Wavelength, an MCP-enabled artifact containing tasks, RAID-log items, and program state. I also built Claude skills and Langfuse evaluation prompts for daily operations and assessment of agent-response quality.
What changed
I helped raise agentic-coding adoption from roughly 2% to 43% of merge requests across the broader transformation work; it was still climbing at departure.
Following the March 2026 roadshow, product managers began producing prototypes of their own.
Related build logs
Context Engineering Is a Soda Gun
Context layers, useful constraints, and evals explained through the drinks people actually order.
The Four Stages Nobody Tells You About
The people getting real value from AI aren't the ones with the best prompts — they're the ones who moved through four predictable stages as fast as possible. A map for getting unstuck from prompt obsession.
The newsletter
Get the build notes.
I write about building AI systems in public—what worked, what broke, and what I’d do differently.