muunai
Making the AI’s draft easier to read, edit, and sign off on
muunai is a Munich-based med-tech startup building an AI documentation tool for doctors. Clinicians record a patient conversation and the app generates a structured medical report they can edit and finalize.
I was one of three designers on a two-month engagement. I owned the redesign of the report page, where doctors review and finalize what the AI model produced. I also built the component library the product didn’t have.
-
Role
- UX/UI Designer
-
Team
-
-
Timeline
- Oct 2024 — Jan 2025
-
Focus
- Information Architecture
- Design Systems
Problem
Doctors couldn’t tell where the report ended and the editing tools began, and the button that sent a finished report to a patient file was an icon in the corner.
We ran task-based sessions with three medical students, asking each to create a report, edit a section, and finalize and send it. Students rather than practicing doctors, partly because doctors were hard to get time with, but also because none of them had seen the app before, which is exactly the first-run experience we were trying to fix. All three struggled in the same three places:
-
Unintuitive layout and navigation
Users didn’t know where to start.
-
Overwhelming Smart Edit feature
Users didn’t fully understand how it worked.
-
Unclear report finalization
Save/send buttons were hard to locate and interpret.
Process
Two things had to happen before I could redesign the report page: figuring out the layout, and building the pieces to make it with.
I wireframed layouts to test one idea: separating what the doctor puts in from what the AI produces, so the two are never visually confused. The same goal shaped the home page, where I defined the usability improvements while another designer built the UI: collapse the sidebar, reveal fields progressively, keep less on screen at any one moment.
muunai had a palette and typography but no component library. I built one, covering button states, card styles, and form layouts, so the redesign stayed consistent with the existing product and the team had something to build on afterward.
Solution
The report page now separates what the doctor puts in from what the AI produces.
Input sits on the left, the generated report on the right, with section dividers marking where one part of the document ends and the next begins. Save and send moved to where the doctor finishes reading rather than the corner they started in.
A report you can move through in order
Section dividers give the report visible structure, so a doctor scanning for the part they need can find it without reading the whole document. The layout follows the order they work in: review, edit, finalize.
Smart Edit where you expect it
The Smart Edit panel moved next to the report section it edits, and narrowed so the report stays the widest thing on screen. The tool assists the document rather than competing with it for attention.
Save and send as buttons, not icons
The two actions that commit a report to a patient file were icons in the top-right corner. They are now labeled buttons at the bottom of the report, where the doctor arrives after reading it.
Where it stands
The redesign wasn’t implemented during our engagement, and we didn’t get to test it with the people who would use it.
Our research ran with three medical students standing in for practicing doctors. They were good proxies for first-run comprehension, since none of them had seen the app before, but they don’t carry a doctor’s time pressure or the weight of signing a clinical document. Validating the redesign would mean putting it in front of working clinicians.
The harder problem went unsolved. We made Smart Edit easier to find and less visually noisy, but a doctor still can’t see at a glance which words the AI wrote and which are their own. In a document they sign their name to, that distinction is the thing worth designing for, and it’s where I’d start if I picked this up again.