An internal GenAI search tool for a large enterprise.
My role & impact
Product Designer, sole designer across the pre-sale prototype and the shipped search tool.
My responsibilities
The brief
It began as a pitch prototype, built to make the client believe the product. It became the full engagement.
The work started as a pre-sale prototype, a demo built for a pitch. A pitch prototype has one job: make the client believe the product. So it has to feel real and run cleanly in a demo, even where the logic behind it is still thin.
The prototype became the full engagement.
The design
A single search surface where a team could query across their internal apps, with prompts to show them what to ask.
What I designed for the shipped tool was a single query surface: a search bar where someone could ask a question and get answers pulled from across the company's internal apps, instead of hunting through each one separately. Around it, a set of suggested prompts and shortcuts to those apps.
The empty state is the real design problem in a tool like this. A blank search bar tells you nothing about what it can do. The suggested prompts and the app shortcuts answer "what can I even ask?" before the user has to work it out for themselves.
Reflections
Pre-sale design and shipped design are different jobs.
The prototype existed to make the client see the product. The shipped tool existed to work. Knowing which one I was building changed what to polish and what to leave rough. In a pitch, the demo has to feel real. In the product, it has to be real.
Measuring adoption Defined adoption for a program with no usage data, and built the screen that reports it. 
Building the operator workspace Sole designer on all three products in a shipped program credited with 10–25% productivity gains. 