altrix

Role

UX Design Intern

Timeline

Mar – Oct 2025

Tools

Figma

Team

engineer

product manager

ux content writer

PROBLEM

Nurses chart later in the shift from memory, not while they’re with patients.

Nurses bounce between patients and get interrupted all the time. Most notes end up scribbled on paper.

Charting happens later in the shift. Anything that didn’t make it onto paper comes from memory, and things slip.

That’s a problem because the chart is the patient’s official record. Billing and compliance rely on it.

A typical workflow for nurses looks like this.

  1. during careCare for patient
  2. between patientsScribble patient info on paper
  3. later in the shiftChart from scraps + memory
  4. next patient

After talking to nurses, a simpler workflow would look like this.

The goal was simple: reduce cognitive load during documentation.

Through interviews with University of Michigan nurses, I explored how AI could support patient care workflows.

  1. during careCare for patient + chart as you go
  2. between patientsReview + confirm AI notes
  3. later in the shiftAlready verified + charted
PROCESS

Three iterations, each shaped by nurse feedback.

I tested prototypes quickly and refined each iteration based on how nurses actually worked.

01. Passive transcription

Captured conversations during care and the AI would parse through the information.

Nurses reviewed generated summaries before anything was committed to the EHR.

NURSE FEEDBACK

“It helped with typing, but I still had to figure out what needed to be charted.”

“I was still relying on memory. The transcription didn’t help with that part.”

TAKEAWAYS

Typing wasn’t the core problem. Memory and recall were.

02. Conversational chatbot

I explored a guided conversational interface that prompted nurses through documentation.

The assistant provided contextual awareness while keeping nurses in the loop.

NURSE FEEDBACK

“I need to choose which patient I’m charting for. I can’t trust the AI to figure that out.”

“I need to know what’s an AI suggestion and what’s going in the official record.”

TAKEAWAYS

Trust meant predictability and control. This shaped the final iteration.

03. Refined assistant

The final direction introduced explicit patient selection and clearer separation between AI suggestions and confirmed records.

NURSE FEEDBACK

“Finally knowing exactly which patient I’m documenting for makes this feel trustworthy.”

TAKEAWAYS

Giving nurses visibility and control over every step made the workflow feel safe enough to integrate into clinical environments.

After selecting a patient, nurses can review, edit, and confirm notes before sending them to the EHR.

Confirming notes before EHR submission

Final in-app experience

Patient selection screen
SELECT PATIENT
Charting patient information screen
CHART PATIENT INFO
Final confirmation screen
FINAL CONFIRMATION
OUTCOME

A nurse-guided AI assistant designed around trust and real clinical workflows.

AI suggests. Nurses decide.

Nothing reaches the EHR without explicit confirmation.

The final design reduced documentation friction and preserved the oversight required in healthcare.

~17%

Of each shift estimated to go back to nurses, away from charting

$220k+

Pre-seed funding secured after acceptance into Techstars

100+

Hospitals involved in pilot discussions through HCA Health’s network