A clay coin-operated viewer on a small green hill, the kind that looks out over San Francisco.

Building an internal AI research assistant

As the head of product research at Jeeves, I needed to ensure that we were operating at the edge of AI capabilities and transforming how our internal teams were consuming insights from our user, market, and competitor research. I accomplished this by building a standalone AI web app that any Jeeves team member could access and self-serve their research needs. The result was 25% company-wide adoption in its first 30 days and a 75% reduction in research team requests, allowing our researchers to focus on bigger projects while those who needed customer, market, or competitor insights could easily find them at any hour of the day or night.

  1. Research & Discovery

    As a new function, the research team was overwhelmed with one-off requests for historical research reports, insights, and quotes.

  2. Design & Prototype

    I worked with internal ops, GTM, and product team members to identify what kind of enumerated, temporal, and analytical questions they most often wanted to answer, but couldn’t. I quickly built a prototype using Cursor and later Claude Code and iterated on the user experience that would best allow them to self-serve their research needs.

  3. Ship & Learn

    I launched the research assistant at an all-hands, resulting in a 25% adoption rate across the company. I learned that interacting with the research assistant inside of Slack would allow team members to reach insights where they were already working.