Draft doctor’s letter from bullet points
From treatment bullet points comes a drafted letter — strictly on-premise.
Local LLMThe problem
Doctor’s letters cost an enormous amount of time — time that is then missing with patients. Cloud AI is off-limits here: this is highly sensitive health data.
The solution
The doctor dictates or types bullet points; a locally running model drafts a structured letter in the practice’s usual style. Nothing leaves the practice server. The doctor reviews, corrects, signs.
How it works
- The model runs on-premise on a small GPU machine in the practice.
- Bullet points → structured draft (history, findings, plan).
- The practice’s own text blocks and tone are applied.
- No internet needed — 0 patient data leaves the building.
What changes
Noticeably less documentation time, more time with patients — without scratching data protection in the slightest.
Honest take
This is a case where AI genuinely helps — but ONLY on-premise. A cloud solution would simply be the wrong path here. And the draft stays a draft: medical responsibility is not delegated to a model.
