Designing Trust into AI-Assisted Expert Work
Digitising Wound Care for Elderly Patients
Digitising Wound Care for Elderly Patients
Digitising Wound Care for Elderly Patients
Digitising Wound Care for Elderly Patients
A concept exploration across two shipped 0→1 products — Zonder (healthcare)
Transforming paper-based workflows into an intuitive digital solution that helps clinicians provide better care for elderly patients with chronic ulcer wounds.
TL;DR
Zonder moves elderly clinicians off paper wound charts putting a high-stakes, zero-tolerance-for-error decision in front of a time-pressed expert. Repetition is the point: this isn't a reaction to a trend, it's a framework I now design with by default.
The research turned up the same tension before I ever thought about AI. On Zonder, clinicians over 55 "feared that unfamiliar technology could disrupt their efficiency and undermine their expertise" — they weren't anti-technology, they were protecting a skill they'd spent decades building.
Principles for trustworthy AI
Draft, don't decide
Show the seam
Fail loud, not silent
Zonder: AI-assisted wound assessment
Zonder is a digital wound-assessment tool that replaced a 1.5-hour paper process for clinicians treating elderly patients with chronic ulcers. What shipped: documentation time dropped to an average of ~14 minutes for 17 of 22 clinicians, 92% preferred it to the old method, and patient throughput rose 59% eight months post-launch.
The AI twist: clinicians are still hand-documenting mid-treatment and that's exactly the moment they can least afford to look at a screen. So I explored what happens if the visit itself becomes the input: AI listens and drafts the assessment, the clinician reviews and confirms before anything saves. Three principles guided every screen below: draft, don't decide; show the seam; fail loud, not silent.
Where AI could take this:
Entry point
Clinicians can switch to AI-assisted recording anytime
Show the seam during success
AI-suggested answers are confirmed or overridden by the clinician
Fail loud, not silent in visual interface
When AI isn't confident in a photo, it flags it for review instead of guessing
Draft, don't decide
Recording is in progress & nothing is saved yet
Fail loud, not silent in voice interface
When AI can't confidently answer, it says so instead of guessing
Next steps (if I cont'd)
Open question: does drafting from ambient audio reduce documentation time further, or does review-and-correct simply replace typing time with editing time? That's a usability study, not an assumption.
