Medical AI scribe

A medical AI scribe is a software tool that listens to a clinical consultation, transcribes what is said, and produces a structured clinic note¹ — typically organised under standard clinical headings (history, examination, assessment, plan) — that the clinician then reviews, edits, and signs off. The clinician remains the author of the clinical record; the AI scribe does the typing. At Breastory, the AI scribe used by Dr Fiona Tsang-Wright is Heidi, a clinical-grade tool widely deployed across UK NHS and private practice in 2026.

The practical effect for a patient is more eye contact during the consultation, less time watching the consultant type. The conversation is the same; what changes is who handles the note-taking afterwards. The clinical note that lands in your record and goes to your GP is the consultant’s note — they review and edit what the AI produces before signing it off. You can opt out of AI scribing at the start of any appointment, and the consultant will take notes by hand for that visit. See AI in your care for the practice’s overall approach.

Related terms: AI in your care · AI-assisted radiology · Multidisciplinary team · Triple assessment

Definition What it is, and what it does

A medical AI scribe is a software application that:

1. Listens to a consultation through a microphone (typically the consultant’s laptop or tablet in the consulting room).
2. Transcribes the conversation in real time using speech-recognition technology trained on medical vocabulary.
3. Generates a structured clinical note by analysing the transcript — usually organised under standard headings such as History of presenting complaint, Past medical history, Examination findings, Discussion, Plan — and putting the right pieces of the conversation under the right heading.
4. Presents the draft note to the clinician at the end of the consultation (or shortly afterwards).
5. Waits for the clinician to review, edit, and sign off the note before it becomes part of the patient’s clinical record.

The point of the tool is workflow time. Clinicians traditionally spent a significant proportion of every consultation typing notes, or wrote them up afterwards from memory or scribbled notes — neither of which produces an ideal record. An AI scribe shifts that burden to software, freeing the clinician to focus on the conversation in the moment and to review a structured draft afterwards.

It does not replace the clinician’s judgment. The note is the clinician’s note, and they are accountable for what it says — the AI is a drafting tool, not a colleague.

How it differs How it differs from older voice-recognition systems

Voice-recognition transcription has existed in clinical workflows since the 1990s — most consultants have, at some point, used Dragon Medical or a similar tool to dictate letters and notes. Those systems transcribed verbatim what the clinician dictated; they didn’t structure the content.

The current generation of AI scribes (introduced from around 2022 onwards, accelerated by large-language-model technology)⁴ does something different — it listens to the two-way conversation between clinician and patient, then structures what was said into a clinical note. This requires the AI to:

• Distinguish clinician speech from patient speech.
• Identify which parts of the conversation are relevant to which section of the note.
• Generate clinical-quality language — appropriate medical terminology, correct grammar, a register consistent with how clinicians write.
• Filter out irrelevant content (small talk, interruptions).

This is the technology shift that has made AI scribing widely useful — and widely deployed — over the last 2 to 3 years. Heidi⁴ is one of the leading UK and Australian-origin tools in this space, alongside others like Nabla, Suki, Abridge, and Augmedix.

At Breastory Heidi specifically — what’s used at Breastory

Heidi is a clinical-grade AI scribe used by Dr Fiona Tsang-Wright at consultations. Things worth knowing:

• Designed for clinical use. Heidi is built specifically for healthcare consultations — its transcription, vocabulary, and output structures are clinical, not generic. It is registered with the MHRA as a Class I medical device for clinical summarisation.
• UK and Commonwealth deployment. Heidi is used in UK NHS and private practice as well as in Australia, New Zealand, and the US⁴. The scale of deployment matters because it shapes the data-handling guarantees the company can reasonably make.
• Transparent data handling. Heidi publishes its data-retention policy, its data-processing locations, and the specific UK GDPR-relevant guarantees³⁴. The current published policy at the time of writing retains the audio for a defined period typically within 7 days (often deleted soon after the note is verified) before deletion.
• Clinician remains the author. The note Heidi produces is a draft. Fiona reviews it, edits it for accuracy and clinical appropriateness, and signs it off. The note that lands in your record is hers, not the AI’s.
• No autonomous decision-making. Heidi does not generate a diagnosis or a treatment plan. It produces the note that documents the consultation; the diagnosis and plan are the clinician’s, formed through the same clinical reasoning that has always applied.

Data handling What happens to the audio

This is the part patients most want to know, so it’s spelled out in detail.

1. Audio is captured during the consultation through the device microphone in the consulting room.
2. Audio is processed to generate the transcript and the structured clinical note. The processing happens on Heidi’s infrastructure — typically a cloud service operated within UK / EU data-protection jurisdiction.
3. Transcript and note are returned to the consultant’s interface.
4. Audio is retained for a defined period — typically within 7 days at the time of writing⁴ (often deleted once the verified note is signed off) — and then deleted by Heidi’s standard data-retention policy.
5. Transcript and structured note live in your clinical record at the practice, with the same access controls as any other clinical note (the consultant; the practice PA for administrative purposes; your GP if a copy is sent at your request).

If you want the audio deleted earlier than the standard retention period, the practice can request that on your behalf — contact Sarah at [email protected].

Transparency and opt-out How Heidi is introduced at Breastory

At the start of a consultation where Heidi is being used, Fiona will mention it briefly and explain that you can ask for hand-written notes instead. Under NHS England ambient-scribe guidance and UK GDPR, explicit consent is not required for individual care — the obligation is transparency and the opportunity to opt out.

Your choice applies to that consultation. If you are comfortable with Heidi at one appointment, you can decline at the next — or vice versa.

If you decline, Fiona will take notes by hand. The clinical record is the same — the note in your file at the end of the appointment is a clinician’s note either way; what changes is whether the AI drafted it first.

Your options What you can ask

• “Don’t use Heidi for this appointment.” Fine. Fiona writes a hand-written note instead. No need to justify it and it does not affect your care².
• “Delete the audio earlier than the standard retention window.” The practice can request early deletion on your behalf — contact Sarah at [email protected]⁴.
• “Who can read the transcript?” The same people who can read your other clinical notes — Fiona, Sarah for administrative purposes, and your GP if a copy of the consultation letter is sent. Heidi’s own staff do not routinely access individual transcripts⁴.
• “Where is my consultation audio stored?” On Heidi’s infrastructure, typically cloud servers in the UK or EU, operated under Heidi’s published data-handling policy³⁴.
• “Can I have a copy of the note?” Yes. The clinical letter that follows the consultation is the patient-facing version of the note; a fuller copy can be requested through the practice.

At consultation What to discuss with Fiona

If AI scribing matters to you, you don’t need to raise it in advance — Fiona will mention it at the start of the consultation and confirm you’re comfortable. If you want to think it through before the appointment, useful things to consider:

• Whether you’d prefer hand-written notes for sensitive parts of the conversation. It is fine to opt out part-way through; Fiona can pause Heidi if asked.
• Whether you want a copy of the consultation letter sent to your GP — the standard at Breastory is to send a letter to your GP unless you ask otherwise.
• Any specific data-handling questions — Sarah can route these to Heidi’s data protection contact if needed.

Resources Further reading

• AI in your care — the practice’s overall approach to AI tools used in imaging and consultation.
• Privacy notice — how the practice handles personal and clinical data.
• AI-assisted radiology — the imaging-side counterpart to AI scribing.
• Heidi Health — vendor documentation, data-handling policy, sub-processor list.
• GMC Good Medical Practice — the duty of openness and accountability that applies regardless of the tools a clinician uses.
• ICO UK GDPR guidance for health and social care — the UK data-protection framework.