Ambient documentation
The AI scribe: listen to the visit, draft the note — attacking healthcare's documentation burden at its source.
In one line
Ambient documentation systems capture the clinician–patient conversation (with consent), transcribe it, and draft the clinical note — returning eye contact to the exam room and evenings to the clinician.
How it works
A pipeline: speech-to-text tuned for medical vocabulary and multiple speakers → an LLM restructures dialogue into note sections (HPI, exam, assessment, plan) → optional coded outputs (problems, orders) → and the non-negotiable step, clinician review and signature: the human owns the record; the machine drafts it. Quality hinges on audio capture, accent and code-switching robustness (Hindi-English consultations are a real test), hallucination control, and specialty fit.
Where it shows up in digital health
The fastest-growing clinical-AI category in deployment — major EHRs ship integrated scribes, and burnout metrics are the headline results. For informaticians the questions that matter: where does audio go and for how long (consent and DPDP/HIPAA), how are errors measured post-deployment, and does the draft note nudge clinical reasoning? "Drafts, never decides" is the pattern to hold onto.