Ask any clinician what they would do with an extra hour each day. Almost none will say more documentation. Yet documentation is exactly where the hours go. The after-hours charting, and the screen time that should have been eye contact. This is the burden ambient AI scribes are built to lift.

An ambient scribe listens to the consultation as it happens. It runs speech recognition and language models over the conversation. Then it drafts a structured clinical note, often in a familiar SOAP layout, ready for the clinician to review. It does not diagnose or suggest treatment. It captures what was said and shapes it into a record.

The pull is obvious once you see the numbers. In a quality-improvement study across six United States health systems, ambulatory clinician burnout fell from 51.9 percent to 38.8 percent after thirty days with an ambient scribe. At The Permanente Medical Group, the tool saved an estimated 15,791 documentation hours across roughly 2.5 million encounters in its first year. One narrative review found after-hours charting dropping from 50.6 minutes to 35.4 minutes a day. Patients notice too. Nearly half reported their doctor spent less time looking at the computer.

Where it goes wrong

The same review found something sobering. Omissions made up 86.3 percent of all errors. The model quietly left out clinical detail that was actually said. Those are the dangerous mistakes. They are hard to spot, because catching them means remembering a conversation from hours ago.

Then there is fabrication. Models invent plausible detail that was never discussed. One study saw a scribe state that a patient's weight loss was intentional when nothing of the sort was said. Australian guidance warns about scribes confusing positive and negative findings, or generating examination findings that never happened. These tools are also non-deterministic. When researchers fed the same transcript to a model three times, they got three different notes. Only 52.9 percent of data elements were reported consistently across all three.

This is why the note is a draft and the clinician is the author. The responsibility for the final record does not move. Read every note properly before it enters the chart. Treat the AI as a fast junior who is sometimes confidently wrong.

Doing it properly in Australia

The RACGP is clear that the GP remains responsible for ensuring health records are accurate and up to date. Consent comes first. Get the patient's agreement at the start of each consultation and give them a real chance to decline. Recording a conversation without consent can be a criminal offence under state and territory listening-devices laws, so this is a serious obligation. Verbal consent is generally enough if you document it, though some medical defence organisations want it in writing. Ask yours.

Privacy is the other pillar. Under the Australian Privacy Principles you need to know where the audio and notes are stored and whether the vendor reuses your data. An Australian-based vendor storing data onshore makes this far simpler. AHPRA has been blunt on the principle that matters most. There must be human input into every AI output. The Australian Commission on Safety and Quality in Health Care has even published an ambient-scribe safety scenario to help services think this through. Wrap the tool in real governance. Write a policy and audit the notes.

One Australian team worth a mention here is Heidi Health. Founded in Melbourne in 2021, Heidi launched its scribe in February 2024 and now supports millions of consultations a week. In October 2025 it raised a USD 65 million Series B led by Point72. It holds SOC 2 Type 2 and ISO 27001. It integrates with Best Practice and MediRecords. The compliance posture is the kind of detail to look for, whichever vendor you assess.

Ambient scribing is one of the few digital health tools where clinicians feel the benefit in the first week. The question is whether your service is set up to capture that benefit safely. Have you decided who signs off the note and where your patients' data lives?

If you are weighing up ambient scribing for your practice or health service, talk to Coterie Health. We help Australian teams adopt clinical AI with the governance and the clinical judgement it needs.