Scaling AI product development across Copilot
Led the Data Intelligence Platform supporting M365 Copilot — turning repeated, bespoke AI work into a self-serve capability that many product teams could build on.
reduction in time to deploy AI agents
increase in experimentation agility
Copilot teams adopting the capability
MAU served across Copilot experiences
Context
Across Copilot, teams were solving the same problems independently: getting enterprise data into a usable shape, grounding responses in the right context, standing up an agent, then proving it worked.
Every team paid that cost again, and the cost was mostly invisible — it showed up as slow experimentation rather than as a line item.
Visual model
Data Intelligence Platform
Self-serve agent creation, grounding and evaluation
- Copilot agent experiences
- Contextual grounding
- Personalization
- Enterprise data workflows
15+ Copilot teams building on one capability · 600M+ MAU served
One platform capability enabling many downstream AI products and teams.
What I focused on
- 01
Treated internal product teams as the customer, with adoption as the success metric rather than shipped surface area.
- 02
Made agent creation and contextual grounding self-serve, so teams could move without a dependency on mine.
- 03
Prioritized the parts of the workflow that blocked experimentation first — speed of iteration compounds faster than features.
- 04
Held a single quality and measurement bar across the platform, so downstream teams inherited it rather than reinventing it.
Why platform thinking mattered here
A capability used by one product is a feature. The same capability adopted by fifteen teams changes how an organisation builds — it lowers the cost of trying something, which changes what teams are willing to try.
The measurable outcome was not the platform itself. It was that AI product development across Copilot got faster, more consistent and easier to evaluate.
Themes
- Enterprise data workflows
- AI agent creation
- Personalization & contextual grounding
- Platform thinking
- Self-serve product experiences
- Cross-team adoption