← All impact
ZillowAI Enablement LeadDec 2025 – now
AI adoption across 2,700+ people
Getting Product, Engineering, Marketing and Design to use AI well, and proving it with telemetry instead of self-report.
2,700+
people in my remit across four orgs
>90%
completion on self-paced AI training
2+ hrs
of engagement on that training
The situation
The AI tools were already there. What held people back was onboarding friction and not knowing what the tools were good for. Training was measured by views and star ratings, which can't tell you whether anyone's work changed.
What I did
- Wrote the AI-enablement operating model: intake, prioritization, content cadence, and adoption metrics reported to VP+ leadership.
- Built and run an AI Champions network across the largest orgs, publishing winning workflows back to the business every two weeks.
- Built a self-teaching training harness where the AI tool teaches its own use (Claude Code teaching Claude Code). Champion-designed fluency rubrics grade the learner, a capstone proves the skill, and telemetry confirms real workflows were run.
- Built the AI Enablement Hub, the front door for learning and using AI, with usage telemetry piped to the data platform so capability is measured in the product.
- Consult on and champion the internal agent platform and a marketplace for publishing skills and MCP servers.
- Teach hands-on across ChatGPT and Codex, Claude Code, Cursor and MCP, including cost, context and data-handling practice.
What changed
- >90% completion and 2+ hours of engagement on self-paced training, in a format that checks what people actually did.
- Training content is moving out of video and into in-tool harnesses, because they produce evidence of skill rather than view counts.
- Rollouts linked to faster product and software delivery cycles.
Claude CodeChatGPT & CodexCursorMCPAgent harnessesLearning telemetry
