Designing once, scaling across the whole suite

Designed a progressive disclosure system that became the foundational pattern for how educators access data across multiple Microsoft learning products. The challenge: balance quick insights with deep exploration across different contexts.

Interviews with teachers revealed they care about outlier data—a student who usually scores 90% but suddenly drops 20 points. The grade wasn't the insight; the story behind it was. That finding shaped the entire progressive disclosure system: quick surface-level view for scanning, deep context available on demand.

Early ideas ranged from info-icons to card flips to color coding. But Microsoft couldn't universally define what counts as "good" or "bad" trend across every metric—so those approaches fell away. I audited every insight card across every Learning Accelerator and built a framework to ensure the pattern would look and behave consistently everywhere it appeared.

Teachers usually land on a review screen mid-grading. A hard jump to a separate Insights app is disorienting. So the pattern steps through three levels instead—card, modal, full app—keeping teachers in flow. On desktop it's a click; on iPad it's a tap. Same behavior everywhere.

EDU Copilot had to work consistently across the entire M365 suite: Word, PowerPoint, Excel, Outlook, Teams. That meant partnering with every product team and establishing a shared direction. The hard part wasn't building Copilot. It was ensuring it behaved the same way everywhere teachers encountered it.

Interviews with educators revealed the constraint wasn't technical. It was human. How much did teachers trust an AI suggestion enough to act on it? Which problems did they actually want help with? That trust gap shaped everything.

Built an experience framework for how and where Copilot appears across Education products, and a way to prioritize features based on where educators' trust actually was, not where the technology could go.

I explored multiple interaction patterns: in-context experiences (hover states, right-click selections) versus a dedicated chat pane. Both approaches were tested to understand which felt more natural and trustworthy. AI-assisted rubrics shipped first, not because it was technically easier, but as a proof point. It matched what teachers were ready for, rather than asking for more trust than they had.