GenFEI

GenFEI is an AI-powered chatbot built for the Research & Engineering team at Kimberly-Clark. It connects to multiple internal knowledge bases and allows highly technical users to query everything from product data to dense engineering documentation in natural language.

This was our first serious attempt at building an internal AI product—and it set the design and usability standard for everything that followed.

The problem

Before GenFEI, finding information meant knowing exactly where to look. Data lived in different systems, with different search experiences, and different mental models. Engineers and researchers often had to guess which system might contain the answer they needed—and if they guessed wrong, they’d start over somewhere else.

These users were deeply technical, but their time was limited. They didn’t want to hunt for documentation—they wanted answers. The challenge wasn’t just building AI logic. It was designing an interface that made a very complex system feel simple, predictable, and trustworthy, even across multi-step, technical conversations.

We also had to deal with new questions no one had answered before: How much control should users have over AI behavior? How do you show advanced options without overwhelming people? How do you keep context across long, technical conversations?

My role

I owned UX research, UI/UX design, and frontend development. I worked closely with engineering to translate system complexity into something humans could actually use.

Designing the solution

I designed and built a conversational interface focused on clarity, flexibility, and control. The goal was to make the system powerful without making it feel complicated.

One key feature was dynamic follow-up suggestions. Based on what a user asked, the interface suggested relevant next questions so users could explore a topic without constantly rephrasing or guessing what to ask next.

Another important feature was temperature control. Instead of hiding AI behavior behind a black box, I exposed it in a simple way. Users could choose whether they wanted strict, precise answers for validation or more creative, high-level thinking for ideation. It gave users agency over how the AI behaved, which built trust.

Under the hood, the system routed questions to different knowledge bases. On the frontend, my job was to make sure users never had to think about that complexity. The interface used progressive disclosure, showing advanced controls only when people needed them.

I also built in feedback loops. Users could quickly downvote bad answers, and we ran office hours to collect deeper feedback. That input directly shaped iteration.

Building the experience

I built the frontend in Next.js with Tailwind and Framer Motion. I used subtle animations to make the experience feel responsive and alive—animated input states, smooth message loading, and gentle transitions between states.

These micro-interactions weren’t decoration. They helped users understand what the system was doing and reduced friction in long, multi-turn conversations. When you’re asking complex questions, speed and clarity matter.

The result

GenFEI launched as the first chatbot of its kind internally—and it immediately felt different from typical internal tools. Users commented on how polished it felt, more like a customer-facing product than an internal prototype.

More importantly, it worked. People could finally ask natural-language questions and get answers without worrying about where the data lived. It gained quick traction and became the design and usability reference for future AI tools at Kimberly-Clark.

This project taught me how much UX matters in AI products. The intelligence doesn’t matter if people don’t trust it, understand it, or enjoy using it. GenFEI showed that even highly technical tools can—and should—feel thoughtful, human, and well-designed.