Bringing 1,200+ voices into an AI chatbot
As the Assoc. Director of Product Design, I initiated and led a research partnership that gave Roo, an AI chatbot with more than 5 million conversations, a clear direction. I set it up to work for both sides. University of Georgia researcher Divya S. gained data for her academic work. Planned Parenthood gained a 1,200+ person study for $15K, the cost of recruitment, which I found in our existing budget.
5M+
Conversations on Roo since launch
$15K
Research budget, found inside existing funds
1,200+
Total research participants
3
Research publications
Roo had already held millions of conversations about sex ed, but no one agreed on how it should evolve. Meanwhile, health orgs around us were racing to add and optimize AI tools. It was tempting to keep pace, but I knew adding features without knowing what problems we were solving put users' trust at risk. We needed to hear from them first.
A national study usually needs its own budget. Instead, I built a partnership that worked for both sides. Divya S., a PhD researcher at the University of Georgia, gained data for her academic work. We gained a 1,200+ person study for $15K, the cost of recruitment. I found the money in unused funds within our own budget.
Every step went through IRB approval and Planned Parenthood’s legal and security review, which slowed our start. To speed things up, I brought in Nicole Levitz, Director of Digital Health Interventions & Insights, who had led a similar partnership before. Her experience helped us move faster, and her guidance helped me avoid common pitfalls.
Much of my role became translation: reviewing Divya’s research and matching academic timelines and priorities to product needs. As findings came in, I saw an opportunity for Elise Gust, a product designer on my team, to step into greater leadership. She led ideation and prototyping while I provided feedback and direction. She went on to co-author 3 research publications with Divya and was promoted shortly after.
Elise’s concepts went into usability testing, and the strongest carried into our roadmap. These included privacy indicators to build trust with users and exploring alternative education formats. The research gave the team an evidence-based direction for Roo. Because it was published, the findings now reach beyond Planned Parenthood and add to public knowledge on AI in digital health.
