The client
A team building a dental software product who wanted to add AI directly into the clinical workflow their users already lived in — not as a separate tool clinicians had to leave the chart to use.
The challenge
Treatment planning and imaging review are where a lot of clinical time goes, and where inconsistency creeps in between providers. The team wanted AI to take the first pass — but in a way clinicians could trust, correct, and own.
- Drafting plans by hand took time, and the structure varied from provider to provider.
- Imaging review meant pulling up images separately and reading them without any structured assist.
- Any AI had to be clinician-in-the-loop — a draft to review and edit, never an unsupervised decision.
- It had to live inside the existing dental product, fitting the data and the workflow already in place.
What we built
An AI module embedded in the product, designed so the clinician stays in control at every step:
- AI-drafted treatment plans generated from the patient's clinical data, presented as an editable starting point the clinician reviews and finalizes.
- Dental image processing that surfaces findings to support the clinician's read, rather than replacing it.
- A clinician-in-the-loop design throughout — every AI output is framed as a suggestion, with clear review and override before anything is committed.
- Native integration into the dental product, so the assist appears where the work already happens and writes into the existing flow.
We started in Figma, mapped the real clinical workflow first, and built the AI to fit how clinicians actually think and work — with their feedback shaping each iteration.
The results
- Treatment-plan drafting became a review-and-refine step rather than a blank-page task, helping plans come together more consistently.
- Imaging review gained a structured first pass the clinician could lean on or set aside.
- Clinicians kept full authority over every plan and read — the AI assisted, it never decided.
- The capability landed inside the product users already trusted, with no separate tool to learn.
Why it worked
The module earned trust because it respected the clinician. AI did the tedious first pass, the clinician made the call, and everything happened inside the workflow they already knew. Built as an ongoing partnership — designed around the clinical reality, not bolted on after the fact.