Envisioning the future of search architecture

During my time at Microsoft AI working on the future of Bing, one project I’m proudest of is envisioning the future of search architecture. At the time of my audit, I saw Bing users struggling with inconsistent answer layouts, uneven information density, sporadic “AI-powered” answers, and unpredictable content responses. Within a single session, people often had to filter through cluttered content and repeatedly adjust their query just to get a stable, trustworthy result. For linked experiences and answer cards, Bing was effectively shape-shifting: similar intents produced dozens of different card designs, patterns, and tones. That inconsistency doesn’t just create friction, it erodes users’ confidence in what they’re seeing and whether Bing understands them.

My goal in this work was to shift the experience toward clarity and simplification as a strategy for rebuilding trust. I wanted users to spend less effort parsing the UI and guessing which answer to believe, and more effort evaluating the information itself, knowing it would be presented in familiar, reliable ways.

To ground this work, my role involved a systematic audit across Bing’s core segments. I treated it as both interaction research and content strategy: breaking full results pages down into individual “answer slices” so I could examine each unit in isolation. For every slice, I cataloged design variants, inconsistencies, component usage, tone of voice, and content hierarchy to understand how the system was signaling (or undermining) authority and reliability.

This audit became the foundation for future migration and consolidation efforts. The analysis I drove surfaced where we could normalize answer types, align content patterns around clearer promises, reduce redundant or conflicting treatments, and move toward a more coherent search architecture that helps users recognize, interpret, and ultimately trust Bing’s answers.