SMART NOTES
Fifteen minutes of chart-reading, turned into one AI dashboard
At Mayo Clinic, oncologists rebuilt a patient's story from a static document before every consult — flipping between labs, scans, and notes that never lined up. We redesigned that into an AI-assisted Smart Note anchored to a single timeline.
ROLE
UX Research & Design
DURATION
Jan - May 2025
TEAM
2 Designers
TOOLS
Figma, FigJam, ChatGPT
01- PROBLEM
A 15-minute ritual before every patient walked into a consultation room.
Physicians relied on a manually assembled document to piece together a patient's history before oncology consults — slow, inconsistent, and easy to get wrong under time pressure.
1
Time away from patients
"I would love to use that time with my patients"
2
Text heavy notes
"Content is cluttered with no visual hierarchy"
3
Missing chain of trust
"Not certain if the summary is up to date"
02 - RESEARCH
We started in the hospital, not the file.
I shadowed a full clinic day.
Reviewing physicians go over patient notes before a consult.
Watching a physician take notes live during one.
Adding post-consult updates to a Word doc afterward.
That's the actual workflow we were redesigning.
Affinity Mapping
My teammate and I built a FigJam board to sort raw observations into who was affected, what signaled the problem, and why it was happening, before we let ourselves jump to solutions.
Output → A problem statement
"Oncologists don't lack information — they lack a way to trust it fast enough to act on."
Contextual interviews & co-ideation
We brought layout directions back to the same oncologists rather than testing cold — pressure-testing what would actually survive inside a real consult before committing to high fidelity.
Output → Two wireframe directions, narrowed to a single one with physician input
Usability testing
Once the prototype was clickable, sessions changed from "what do you think of this layout" to "find this lab result, compare these two scans, place this order" — timed, task-based, and run by the same oncologists who'd shaped the wireframes.
Output → Specific friction points, not just a layout people liked
03 - DESIGN DECISIONS
Two frictions, 6 decisions
Both screens below exist because of a specific moment of friction we watched happen. Here's the reasoning behind each one.
1. History was chronological in real life, scattered on the chart
01
Timeline of the oncology history
INSIGHT
​Order matches what physicians already reconstruct in their head, so all data is visually connecting.
03
Toggle switch between avatar and list view
INSIGHT
Lets a physician choose their interface incase they would like to compare the whole history.
02
An avatar view
INSIGHT
Location and history read as one fact instead of two things to mentally connect.
2. "What changed?" meant re-reading the whole chart
04
AI drafted notes that surface since last visit
INSIGHT
Answers the first question of every visit without a full re-read.
06
Built-in notes editor
INSIGHT
Keeps the physician as the final authority, not the AI.
05
A chatbot with starter questions pulled from patient history
INSIGHT
Surfaces context a physician would otherwise have to hold in memory
04 - WALKTHROUGH
See it in motion
The timeline, the AI note, and inline ordering — one pass through a consult, the way an oncologist would actually use it.
REFLECTION
The AI's value came not just from its ability to summarize, but also because it matched how oncologists already think — chronologically, mapped to the body, like a smart note.
