An AI scribe in
general practice

General practice runs on short, booked consults, and a single visit rarely holds a single problem. A patient turns up for a repeat prescription and, on the way out the door, mentions a new symptom, then asks about a chronic disease review that's overdue. The Medicare Benefits Schedule expects a contemporaneous, clinically relevant record of whatever was actually addressed in that slot, not a generic note applied afterward.

The result is a familiar trade-off. Type while the patient talks and the conversation loses its rhythm. Wait until the room is empty and the detail starts to fade, or let the notes pile up until after the last patient leaves. An ambient scribe is built for exactly this gap: it drafts the note from the conversation itself, so the doctor's attention stays on the patient in the room rather than on the keyboard, and the file still gets finished to a proper standard rather than a rushed one.

With the patient's consent given before anything is recorded, a microphone listens to the natural back and forth of the consult rather than a dictated summary read out afterward. The audio is transcribed, and once the visit ends a first draft of the structured note appears in a form the doctor can read, edit and check against what was actually said in the room. If a patient would rather not be recorded, the consult goes ahead as normal and the note gets typed the usual way; consent is a precondition, not a formality worked around.

The same underlying mechanics apply whether the room is general practice, physiotherapy or psychiatry, though the shape of the note and the detail that matters shifts by discipline. See how the approach lines up across specialties for the groundwork this piece builds on.

A consult that covers a medication review, a skin check and a mental health question in ten minutes still needs each problem documented on its own terms: history, examination findings, assessment and plan, in the order a future reader of the file will expect, not the order the conversation actually happened in.

An ambient scribe follows the conversation as it happens and organises the draft by problem rather than requiring the doctor to reconstruct that structure from memory once the patient has left. It's a drafting aid, not a diagnostic one: it does not decide what the assessment should say, only reflects what was discussed, in the wording the doctor used. A visit with three problems produces a longer draft than a visit with one, but the structure underneath, one section per problem, doesn't change with the count.

Chronic disease management and mental health care plans carry a documentation standard that's easy to lose in a busy list, and RACGP standards for general practices set a clear expectation around what a proper record looks like. A patient managing diabetes, hypertension and depression in one visit still needs a distinct plan for each condition, not a single paragraph that blurs the three together and leaves the next reviewer guessing which goal belongs to which problem.

The same applies to a referral letter or a discharge summary: the specialist reading it needs the actual reason for referral and the relevant history, not a diagnosis code with no context. An ambient scribe drafts these documents from what was discussed in the room, though the doctor still confirms clinical detail, any specific requirement of the receiving service, and that the letter reads as their own clinical reasoning before it goes out under their name.

The real test of an ambient scribe isn't whether the transcription is word perfect. It's whether the doctor can open the draft, check it against what they remember of the consult, correct it and sign it off in the couple of minutes before the next patient is called in.

This is where the time saving actually lives, or doesn't. Treat the draft as a rubber stamp and the record is only as good as the transcription happened to be that day. Treat it as a colleague's first pass, worth reading properly rather than skimming, and the review takes minutes rather than the hour of after-hours typing it was meant to replace.

• A structured draft lands in the file shortly after the consult ends, organised by problem
• The doctor reads it against their own memory of the conversation, not against the raw audio
• Medications, doses, allergy status and plan details get checked line by line
• The doctor edits directly, then finalises and signs before anything becomes part of the record
• Nothing is filed unsigned, and the draft state before finalising is kept, not silently overwritten

An ambient scribe is a documentation aid, not a medical device, and it does not diagnose or recommend treatment. Findings that come from looking rather than talking, a rash, a gait, a wound, still need the doctor to describe them out loud or add them to the draft afterward. Anything the note implies about urgency or severity needs the doctor's own clinical read, not the transcript's phrasing of it. If the draft gets something wrong, a misheard dose, a symptom attributed to the wrong problem, that's what the review step exists to catch before it becomes a permanent part of the file.

Health information is treated as sensitive information under the Privacy Act 1988 (Cth) and the Australian Privacy Principles, and that status doesn't change because a scribe is doing the typing. Consent, storage and access all matter exactly as they would for any other clinical record, and AHPRA's expectations around record keeping still apply to the finished note, not to the tool that helped draft it.

None of this replaces the practice management system. An ambient scribe drafts the note; it still needs to sit inside whatever the practice already runs, whether that's Best Practice, MedicalDirector or Genie, and the finished note still lives in the same patient file it always has, next to the results, the scripts and the recall list. Most ambient scribes run from a phone or the computer already on the desk, so adopting one is a change to a habit, not a change to the room.

For a practice manager, that's the practical question worth asking early, not whether the transcription is accurate, but how the draft gets from the scribe into the record the practice already relies on, and who is responsible for checking it landed correctly before the patient's next visit.