Healthcare knowledge base
A healthcare software support team was answering the same tickets over and over, while past solutions sat locked in closed tickets. I designed an AI-powered knowledge base that turns resolved tickets into trusted, searchable articles — with a human always in the loop.
Overview
The answer already exists — in a closed ticket nobody can find.
An AI-powered knowledge base for a support team handling thousands of tickets across a dozen medical products. Every resolved ticket held a real answer, but it stayed locked in a closed case. I designed a platform that turns those tickets into structured, searchable articles — with AI drafting the first version and support experts reviewing before anything publishes.
The hard part
A knowledge base lives or dies on trust. In healthcare support, a stale or wrong answer can misguide a clinic. AI could draft articles fast, but nothing could publish unchecked. The challenge was a system fast enough to keep up with ticket volume, yet safe enough that a human expert signs off on every article, with duplicates caught before they multiply.
Research & insights
What the research told me
Agents kept solving problems that had already been solved — the answers were trapped in closed cases.
No one would use a knowledge base unless an expert had verified the answer first.
Similar tickets could spawn conflicting articles, so duplicate detection had to happen before publishing.
How I worked
Learn the support workflow
I interviewed agents and mapped how tickets get resolved to see where knowledge was being lost.
Design the AI-to-human flow
I shaped how AI drafts an article from a ticket and hands it to an expert for review.
Build in trust signals
Confidence scores, similarity checks and validation states let reviewers approve with confidence.
Make it findable
A clean knowledge base with search, categories and an AI assistant so answers surface fast.
Wireframes & iterations
From rough to right
I iterated on the review and editor layouts — how to show AI confidence, similar-article matches and validation checks without burying the article itself.
Final UI
The shipped design
The platform ships as an AI-powered knowledge base with a human in the loop. AI drafts articles from resolved tickets, experts review with full context, and a clean, searchable library makes answers easy to find.
