Conversational AI
I lead strategic and tactical research for Search Live— a voice-powered AI chat tool built into Google Search. My research drove the product from concept to launch, shaping end-to-end design, system response frameworks, and LLM tuning to ensure conversations are useful, effortless, and serving real user needs.
Featured press: Google Keyword blog, The Verge: Real Time AI Voice Chats, EndGadget
The goal
AI tools are becoming the go-to for answering questions. There's a growing need for hands-free, expressive, and context-aware ways to learn and explore information beyond traditional keywords.
How might we redefine how users leverage Search by introducing natural, voice-based conversation as a more intuitive way to search anything about anything.
The approach
Foundational Research & Market Landscape:
Conducted competitive audits and a literature review to map emerging trends in conversational AI, voice interfaces, and user behaviors across leading tools in the market.
Led in-depth interviews with power users to uncover motivations, current workarounds, and unmet needs around voice search, setting the groundwork for strategic differentiation.
Behavioral & Concept Research:
Drove iterative concept and usability testing to refine the voice experience—from query initiation to follow-up—to ensure ergonomic ease, contextual responsiveness, and seamless user journeys.
Investigated how users process voice interactions cognitively and physically: What should be spoken vs shown? What works in public vs private environments?
Identity & Sentiment Discovery:
Facilitated focus groups to explore brand identity, early perceptions, and emotional resonance — informing product tone and positioning within the Google Search AI ecosystem.
Analyzed emergent themes in how users conceptualize AI assistance tools—identifying key delight and frustration points to refine product strategy.
Scaled Evaluation & Launch Readiness:
Ran a large-scale survey with trusted testers to assess perceptions of helpfulness, trust, and tone—key dimensions influencing long-term adoption and satisfaction in LLM-based products.
Delivered high-impact insights to cross-functional teams and executive leadership that directly informed pre-launch polish and post-launch priorities.
The Impact
Identified critical usability challenges: like cognitive overload from voice-only responses, sparking key design shifts like transcript visibility, response verbosity tuning, and UI affordances for input and output control.
Directly informed LLM prompt tuning for response quality, content strategy, and tone, influencing how the model interprets nuanced queries and delivers trustworthy, engaging, human-like responses.
Enabled the Search org to confidently launch product from 0-market, entering the conversational AI space with a voice-first experience that aligned with users’ authentic behaviors and needs from cars to kitchens to classrooms.