Dental clinical documentation AI company
Dentists approve a clinical note drafted during the appointment, not rebuilt after it
A dental documentation product turns the appointment conversation into a structured clinical note. We work on its drafting path and the dentist's review and approval screen, so the note is drafted during the visit and released only after the dentist signs off.
Healthcare & WellnessClinical documentation AIBuilt from zero
- Built for
- Dentists, during and after appointments
- Approval
- The dentist reviews, edits and approves
- Review screen
- Edits, annotations, overrides, version history
- Transcription
- Whisper and Deepgram
The challenge
Dentists spent time after each appointment turning the conversation and their notes into structured clinical documentation.
What we did
A drafting path from live transcription and patient history to a model-drafted note, and a Next.js review screen where the dentist edits, overrides and approves.
The outcome
The note is drafted during the visit, and nothing reaches the patient record until the dentist approves it. The work continues as the product grows.
The starting point
A clinical documentation product for dentists that turns the appointment conversation into structured clinical documentation. Dentists used to spend time after appointments rebuilding the conversation and their notes into that documentation.
What had to change
Documentation needed to be drafted during the appointment instead of reconstructed afterwards. That meant live transcription, the patient's history and a model working together.
What could not break
- No note is released before the dentist reviews, edits and approves it
- The draft has to be ready during the appointment, not after it
- Drafts draw on the patient's history as well as the conversation
The calls we made
We treated the dentist's approval as part of the product, not a policy on top of it.
- 01
Draft, never release
Clinical notes carry the dentist's responsibility, so the model's output stays a draft until the dentist approves it on the review screen.
Ruled outSending AI output to the patient record or the patient without a clinician's sign off.
- 02
A review screen built for changing the draft
The dentist can edit, annotate and override the draft, and version history keeps what the model wrote apart from what the dentist approved.
Ruled outA plain accept or reject on the whole note.
- 03
Draft during the visit
Transcription runs while the appointment happens, so the dentist reviews a draft instead of reconstructing the conversation afterwards.
- 04
Ground drafts in patient history
Retrieving the patient's history gives the model context that the conversation alone does not hold.
How the system fits together
What we built
Review and approval screen
A Next.js web interface where the dentist edits, annotates and overrides the draft, sees its version history and approves it before anything goes to the patient.
Transcription and history
Appointment audio transcribed with Whisper and Deepgram while the visit is under way, with the patient's history retrieved from Athena as context.
Note drafting
LLM orchestration with LangChain that combines the transcript and the retrieved history into a structured clinical note draft, which waits for the dentist's approval.
How it works
Listen
The appointment conversation is transcribed as it happens.
Retrieve
The patient's history is pulled in as context for the draft.
Draft
The model drafts a structured clinical note from the transcript and the history.
Approve
The dentist reviews, edits and approves, and only then is the note released.
A closer look
What changed
Dentists review a note drafted during the visit instead of rebuilding it afterwards, and the approval step keeps every note under the dentist's control.
Under the hood
- Datics role
- Engineering delivered by Datics
- Integrations
- Athena for patient history, and Whisper and Deepgram for transcription.
- Orchestration
- LLM orchestration with LangChain in Python, combining the transcript with the retrieved patient history.
- Approval boundary
- Nothing is released until the dentist reviews, edits and approves the note.
- Deployment
- Services run across AWS and GCP, behind a Next.js front end.
AI and models
Cloud and infrastructure
Apps and APIs
Also in the stack
How it shipped
In the product's own codebase
The work ships in the product's existing stack and process: a Next.js front end, Python services and infrastructure across AWS and GCP.
Approval first
Every change to the drafting path is checked against one rule: a note only leaves the review screen with the dentist's approval.
What this unlocked
The work continues on the documentation product as it grows.
Related work
Have a product with the same kind of constraints?
Tell us what already exists and what you want it to do. We will tell you what it takes.