Last updated: April 2026

The market for AI medical letters in the UK has moved faster in the past two years than in the previous decade. There are now at least six credible AI tools that claim to help consultants produce clinical correspondence more efficiently. Some are backed by tens of millions in venture funding. Others are procurement items on NHS framework agreements. A few are built by clinicians.

If you're a UK consultant — especially in private practice — this landscape is confusing and the stakes are high enough to get it wrong. A clinical letter is a medico-legal document. A hallucinated diagnosis, a misattributed finding, or a note sent before you've reviewed it isn't just embarrassing. It's a complaint.

This article is a practical guide to AI medical letters in the UK context: what the tools actually do, where they differ, what the privacy considerations really are, and why I believe the architectural choices made by different vendors matter more than their marketing.

How Do AI Medical Letter Tools Actually Work?

AI medical letter tools use one of two approaches: ambient recording (a microphone captures the consultation and AI generates a note from the transcript) or structured input (the clinician enters findings and the AI generates a formatted letter). The key difference is whether the AI interprets audio or formats facts you provide.

Not all AI medical letter tools are built the same way. There are two distinct approaches, and understanding them changes how you evaluate every tool in the market.

Approach 1: Ambient Recording (the majority of tools)

Most AI scribe tools — Heidi Health, Tortus AI, Nuance DAX, Accurx Scribe — work by recording the consultation and generating a clinical note from the transcript.

The workflow looks like this:

You start a recording before the patient enters (or at the start of the consultation)
The audio captures the conversation — clinician, patient, sometimes a third party
The audio is processed by AI, transcribed, and then summarised into a structured clinical format
You review the AI-generated output and make any corrections
The corrected note becomes your clinical record or the basis of your letter

This approach has genuine appeal for high-volume settings. An ED consultant seeing 25 patients in a shift, or a GP working through a 20-minute appointment list, genuinely cannot remember every consultation detail by the end of the day. The ambient recording captures everything so nothing is lost.

Approach 2: Structured Input

An alternative approach doesn't record anything. Instead, the clinician enters the clinical findings they want in the letter — diagnosis, examination, investigations, plan — and the AI generates the formatted letter from those inputs.

This is the approach Docyment uses. You tell the AI what happened. It writes the letter. The clinical content comes from you; the AI handles structure, formatting, and language.

The key difference: in structured input, the AI formats facts you've provided. In ambient recording, the AI interprets audio it has captured.

What Are the UK Privacy Rules for AI Medical Letters?

UK GDPR governs all AI medical letter tools. Ambient recording tools require careful attention to patient consent — the ICO notes that consent in healthcare may not always be "freely given" because of the clinician-patient power imbalance. Structured input tools carry a lighter privacy burden because there is no consultation recording to process.

Privacy considerations for AI medical letters in the UK are governed by UK GDPR and the ICO's guidance on health data processing. Here's what actually matters in practice.

Patient consent for recording

If you use an ambient AI scribe, you are recording the consultation. That recording contains patient-identifiable health data — some of the most sensitive category of personal data under UK GDPR.

The ICO's guidance on consent in healthcare settings is worth reading carefully. It notes that consent "may not always be the most appropriate lawful basis" in healthcare, partly because the imbalance of power between clinician and patient means consent cannot always be considered "freely given." This doesn't mean you can't use these tools — it means your legal basis for processing needs to be thought through.

In private practice, the power dynamic is even more pronounced. A patient paying £250 for a private consultation and asked "do you mind if I record our conversation?" is not in the same position as a patient with a chronic condition who needs ongoing NHS care. The commercial relationship creates an additional imbalance.

Practical implication: If you use ambient recording tools, you need a clear data processing notice that patients receive before the consultation, a robust data processing agreement with the vendor, and confidence that you can answer "where is my patient's data stored?" without hesitation.

Data storage and AI training

Some AI tools use clinical interactions to improve their models. Some don't. The distinction matters for NHS clinic letters and is something every UK consultant should check before deploying any tool.

Look for:

UK-based or EU-based data processing (not US-based, unless specific SCCs are in place)
Explicit contractual exclusion of your data from model training
Clear data retention policy — how long is the audio or transcript stored?
Subprocessor disclosure — does the vendor use third-party AI models (OpenAI, Google, etc.) that themselves have data policies?

What structured input changes

With a structured input approach, there's no audio recording to process. You enter clinical findings — the same information you'd dictate to a secretary — and those inputs are used to generate the letter. The privacy surface is similar to dictating to a human secretary, which is a data processing activity every UK consultant already does.

This doesn't eliminate all privacy considerations, but it materially reduces the consent and recording burden.

What Are the Main AI Medical Letter Tools in the UK?

Heidi Health

An Australian AI scribe platform that has expanded aggressively into the UK market. Ambient recording, transcript-based. Produces clinical notes primarily designed for GP-style SOAP format. Pricing starts at approximately £120/user/month for the clinical plan.

Pros: Polished product, broad specialty coverage, simple to set up.

Cons: Output is clinical notes, not UK clinical letters. American-influenced formatting. Recording consent required. Not designed for private practice letter workflow.

Tortus AI

UK-developed, NHS-focused. Tortus is building deep NHS trust integration — they're currently in deployment at several London NHS trusts. Their accuracy data has been independently published, and it's good: but the published hallucination rate of 1.47% (with 44% of those errors classified as clinically significant) is worth understanding before you sign up.

Pros: UK-built, NHS integration pathway, published accuracy data (transparency is good).

Cons: Enterprise pricing (reported £15,000+ per pilot), NHS procurement focus, not designed for individual private practice.

Nuance DAX (Microsoft)

The enterprise incumbent. Used in large US health systems. Now available in the UK. Sophisticated NLP, deep EHR integration (Epic, Cerner).

Pros: Mature technology, extensive validation data.

Cons: Pricing (~£350-500/user/month) makes it viable only at enterprise scale. Output is designed for US EHR systems, not UK clinical letters. Requires IT deployment.

Accurx Scribe

Accurx is deeply embedded in NHS primary care. Their scribe product uses Tandem Health's AI for ambient transcription with write-back to EMIS and SystmOne.

Pros: Integrated into GP workflow, familiar platform for NHS GPs.

Cons: Designed for NHS GP record-keeping, not private practice correspondence. Not available to independent consultants.

Docyment

Full disclosure: I'm a co-founder of Docyment. I've tried to give balanced information about the other tools above; here's the honest case for a different approach.

Docyment uses structured input rather than ambient recording. You enter the clinical findings you want in the letter — in the format you'd give to a secretary — and Docyment generates a UK clinical letter from those inputs. Output follows standard UK clinic letter conventions: addressed to the GP, proper headers, NHS terminology, structured body.

No recording. No consent burden. No hallucinated content from misheard audio. Letters typically done in under 30 seconds.

Try Docyment free at docyment.com

Frequently Asked Questions

Can AI really be trusted with clinical content?

Yes, with appropriate guardrails. If the AI is formatting structured inputs you've provided, the risk is low — you're reviewing a formatted version of your own clinical findings. If it's interpreting audio recordings, the risk is higher. Whatever tool you use, review every letter before it leaves your hands. AI should assist sign-off, not replace it.

What about medico-legal letters?

Medico-legal documentation requires particular care. For insurers, court proceedings, or personal injury assessments, treat AI output as a first draft to be reviewed line-by-line. The standard of accuracy required in medico-legal work is higher than routine clinical correspondence.

Will my NHS trust provide an AI documentation tool?

Possibly, eventually. The NHS AI scribe procurement process is real but slow — DTAC compliance, DCB0129, DSPT — and trusts that are deploying tools are typically focusing on NHS clinical notes, not private practice correspondence. Your private practice is outside NHS procurement. You choose your own tools — and can adopt good technology now rather than waiting 18 months for a framework approval.

What is the difference between an AI scribe and an AI letter generator?

An AI scribe records the consultation and generates a clinical note from the transcript. An AI letter generator takes findings you enter — by typing, dictating, or scanning — and produces a formatted UK clinic letter. Scribes capture what was said; letter generators format what you tell them.

How much does AI medical letter software cost in the UK?

Pricing varies widely. Enterprise ambient scribes (Nuance DAX) cost £350-500/user/month. Mid-range scribes (Heidi) start around £120/month. Structured AI letter tools like Docyment offer a free tier (20 letters/month) and Pro at $25/month early-adopter pricing.

How to Evaluate an AI Medical Letter Tool for UK Practice

Run this checklist before committing:

Does it produce UK clinical letters (addressed to a specific GP, formatted as correspondence) or clinical notes (structured for EHR systems)?
Does it record the consultation, and if so, what is your legal basis for processing that data?
Where is patient data stored and processed — UK/EU, or US?
Is the data used to train AI models? Is there a contractual opt-out?
Can you see the output before it's sent? Is there a clinician review step?
What is the published or measured accuracy/hallucination rate?
Is the pricing structured for individual clinicians or enterprise procurement?
Does it work with your existing workflow, or does it require you to change how you consult?

Should a UK Consultant Use Ambient AI or Structured Input for Medical Letters?

AI medical letters are real, they work, and they can save a UK private practice consultant 1-2 hours per clinic day. The technology has matured enough to be reliable in routine clinical correspondence.

The choice between ambient recording and structured input is not technical — it's about which bottleneck you're actually trying to solve. If you struggle to remember what happened in each consultation, ambient recording makes sense. If you know exactly what you want in the letter but find the writing process slow and frustrating, structured input is faster, safer, and simpler.

AI Medical Letters UK: What Consultants Need to Know Before Choosing a Tool