MoleSafe — 2 hours → 20 minutes imaging workflow
Redesigned the dermatology total-body imaging workflow with agentic AI for the largest US dermatology network. Patient time dropped from two hours to twenty minutes; per-lesion diagnosis collapsed to seven seconds. 3M+ images migrated through the cutover.
- Patient imaging time
- 2hr → 20min
- Per-lesion diagnosis
- 7s
- Images migrated
- 3M+
- Clinics rolled out / month
- 1 → 5
Total-body photography for melanoma screening is one of those workflows where every wasted minute compounds — patients on the table, dermatologists waiting on pixels, screening throughput capped by the slowest step in the chain. The largest US dermatology network was running this workflow at the wrong order of magnitude. Redesigning it meant treating the workflow itself as the product, not the software underneath — every step over a second got tagged for elimination.
Problem
The largest dermatology network in the USA was running teledermatology with significant inefficiencies. Dermatologists reviewing patient images were often in different states than the nurses capturing them, causing delays and logistical drag. SLAs required reports within two days; the outdated system couldn’t consistently meet that bar.
Pain points compounded: an inefficient software platform, excessive nurse-dermatologist back-and-forth, and slow image processing that turned each patient into hours of manual work. The workflow didn’t scale, and the medical professionals using it were frustrated. The competitor we benchmarked against had a faster on-site experience — but we had a deeper AI platform underneath.
Approach
I dedicated a room in the office, set up mannequins, sourced a stopwatch, and physically reproduced the competitor’s workflow end-to-end. Every step that took longer than a second got tagged for elimination or async-ification. Then I mapped the end-to-end workflow of nurses and dermatologists separately, integrating AI-driven features per role.
- For nurses — AI imaging assist.
Blurriness detection, lesion detection, lesion matching, and dermoscopic ring detection wired into the capture loop so quality is checked at the point of capture, not in QA later.
- SmartSnap auto-detection.
AI tool for automatic spot detection on dermoscopic images. The operator doesn’t aim — they capture. The system identifies and frames the lesion automatically.
- Follow Me cross-device sync.
Mobile capture actions sync to web and Apple tvOS in real time. The dermatologist’s review happens on a different device while imaging continues — no waiting, no batching.
- Context-aware shortcuts.
Adjustments based on the image taken, the role of the user, and the device used. The same button does different things in the right context, instead of forcing the user to navigate menus.
- For dermatologists — diagnostic compression.
Algorithm groups images by spot for a comprehensive view (overview + magnified dermoscopic). State-of-the-art AI skin cancer diagnosis with lesion mapping and change detection — overlays images to identify new, missed, or evolving lesions. Diagnosis-submission collapses to 7 seconds.
Customer obsession is field work. You can’t redesign a clinical workflow from a Figma file — you redesign it with a stopwatch, a mannequin, and a willingness to feel every wasted second yourself.
What shipped
- Imaging time: 2hr → 20min.
A ~80% reduction. Achieved by automating quality checks, wiring AI lesion detection into capture, and removing the synchronous nurse-dermatologist hand-off.
- Per-lesion diagnosis: 7 seconds.
Image grouping + diagnostic shortcuts + AI overlay reduces the dermatologist’s decision loop to seconds per lesion.
- 3M+ images migrated.
2D dermatology images mapped into a 3D mesh as part of the cutover. Visualization improved AND the migration de-risked: users saw familiar data in a richer form, reducing change-aversion.
- Clinic rollout: 1 → 5 per month.
Adoption pace went from one clinic per month to five. Pricing tiers restructured to $800 / $500 / $250 by clinic size, making the platform accessible to smaller practices.
- Multi-device coverage.
Mobile capture, web review, Apple TV display — same workflow state across all surfaces via Follow Me sync.
Outcome
Patient throughput
80% faster imaging
From two hours per patient to twenty minutes. The clinic capacity unlock that comes with that change is what made the rest of the rollout possible.
Diagnostic speed
7-second decisions
Dermatologists handle a significantly higher volume of patients per session without compromising accuracy — AI overlays surface what changed, the human decides.
Scale
5 clinics / month
Adoption pace 5× from baseline. Restructured pricing tiers ($800 / $500 / $250) opened access for smaller practices.
Platform halo
DermEngine +40% DAU
Underlying AI imaging platform saw +40% DAU and +30% consults YoY in the surrounding period — the workflow win pulled the rest of the suite forward.
What I’d do differently
- Get a clinical workflow PM into the office, not just a software PM. The stopwatch-and-mannequin trick worked — but a clinically-trained co-PM would have caught the next round of friction (consent flow, biopsy referral) before we shipped.
- Time-stamp every keystroke in beta, not just session time. "20 minutes" is a great headline; the distribution of that time per nurse / per body region is the next optimization target.
- Treat the FDA-readiness narrative as a launch artifact. The MoleSafe redesign improved DermDx’s FDA submission readiness. That deserved its own GTM story, separate from the per-clinic rollout pitch.
Related work
