Last updated: April 2026

The Letter Bottleneck

Every UK consultant knows the feeling. Clinic finishes at 6pm. You've seen twelve patients. Every one of them needs a letter — to their GP, sometimes to their insurer, occasionally a copy for the patient themselves. Proper NHS clinic letters, not quick notes. Formatted, structured, signed.

And somehow, despite everything that has changed in medicine over the past two decades, the way we produce clinical letters has barely shifted. Dictate. Wait. Review. Correct. Sign. Repeat.

I'm a Consultant Urologist. I write a lot of clinic letters. For years I tolerated this cycle because I assumed it was just part of the job. Then I started asking a simple question: why does a letter that I know the contents of, that I've just discussed with the patient, take three days to reach a GP?

This article explains the state of clinical letter writing in UK practice — honestly, without vendor spin — and what the current options actually look like.

Why Do Clinical Letters Take So Long to Reach GPs?

Clinical letters take 1-4 days to reach a GP because the process involves multiple handoffs: you dictate, your secretary transcribes, you review, your secretary sends. Each step adds delay, and the bottleneck is converting your clinical knowledge into a formatted document — not the clinical knowledge itself.

Before looking at clinical letter software, it helps to understand where the time actually goes.

A typical outpatient clinic letter involves several handoffs:

You see the patient and form a clinical impression
You dictate the letter — either into a handheld recorder, directly into a phone app, or (increasingly) via Dragon Medical
Your secretary receives the audio file, types it up, formats it to your preferred template, and adds the patient's details
You receive the typed draft, review it for accuracy, correct any transcription errors, and sign it off
Your secretary (or a practice management system) sends it

At a busy Spire or Nuffield clinic, step 3 might not happen for 24-48 hours. Your secretary might be dealing with three other consultants' backlogs. The letter might not reach the GP until four or five days after the consultation.

That's a long time for a patient waiting to hear about their PSA result, their knee MRI, or their hysteroscopy findings.

The bottleneck isn't your clinical knowledge. You know exactly what you want in that letter the moment the consultation ends. The bottleneck is converting that knowledge into a correctly formatted document.

What Are the Traditional Options for Clinical Letter Software?

1. The Secretary-Dependent Model

Most consultants rely on a medical secretary to type their letters from dictation. This works when the secretary is experienced, fast, and has capacity. It breaks down when they're off sick, on holiday, handling a backlog, or — as many consultants have discovered — leave with 48 hours' notice.

The secretary model also doesn't scale. If you add clinic sessions, you add to their workload. If their workload exceeds capacity, turnaround suffers.

2. Dragon Medical / Speech Recognition Software

Dragon Medical Professional, the market leader in medical speech recognition, converts your spoken words to text in real time. In theory, you dictate into a document and the letter emerges already typed.

In practice, Dragon requires training, discipline, and patience. You still have to dictate the entire letter word-for-word — diagnosis, findings, plan, sign-off, the lot. You're essentially typing by speaking, which is faster than typing but still slower than it should be. Dragon doesn't know what a UK clinical letter template looks like. It doesn't know the format your GPs expect. It gives you text. The structure is still your problem.

3. Practice Management Software (Healthcode, iGPR, Clinic to Cloud)

Practice management systems typically include some letter functionality — patient record integration, template management, basic mail merge. These are genuinely useful for administrative letters (appointment confirmations, invoice summaries) but fall short for detailed clinical correspondence.

The clinical letter template functionality in most practice systems requires you to write the clinical content yourself. They handle the formatting and mail merge; you still dictate or type everything that matters.

What's Changed: AI Clinical Letter Software in 2026

In the past two years, a new category of clinical letter software has emerged. The most widely discussed tools are ambient AI scribes — Heidi Health, Tortus AI, Nuance DAX, and Accurx Scribe, among others. These tools work by recording the consultation, transcribing it, and using AI to generate a clinical note or letter from the transcript.

If you haven't tried any of these, here's what they actually do:

A microphone (usually your phone or a dedicated device) records the consultation
The audio is processed by AI and converted to a transcript
The AI interprets the clinical content and generates a structured note
You review and edit the note before it becomes part of the record

The appeal is obvious. A consultation you didn't have to document manually. Fifteen minutes saved per patient. An end to the dictation backlog.

The reality is more complicated — and as someone who has looked at this closely, I think it's worth being honest about the limitations.

The Problem with Ambient Recording

The ambient AI scribe category has a structural weakness that doesn't get discussed enough: it captures the full, messy flow of a consultation — hypotheticals, differentials being weighed, plans considered then dropped — and the AI must decide what the final clinical picture actually was.

A published accuracy analysis found that 1.47% of AI-generated sentences in ambient clinical documentation tools contain hallucinated content — information not present in the actual consultation. Of those hallucinations, 44% were classified as clinically significant.

For NHS clinic letters — which are medico-legal documents — a 1.47% sentence-level error rate is not trivial. In a typical clinic letter of 20-30 sentences, the probability of at least one hallucinated sentence is real — and across a busy month of letters, fabricated clinical content reaching GPs or insurers becomes a statistical certainty.

There's also the consent question. UK GDPR requires transparency about data processing, and the ICO's guidance is clear that consent in healthcare settings must be freely given. Asking a patient "do you mind if I record our conversation?" is not a neutral question. For some patients — particularly in private practice, where the relationship is commercial as well as clinical — it changes the dynamic.

Finally: most ambient AI tools generate clinical notes in American EHR format (Assessment and Plan, SOAP note structure). UK private practice consultants need letters — addressed to a specific GP, formatted as correspondence, written in the third person following conventions every NHS-trained doctor recognises.

How Does Structured Input Generate Letters Without Recording?

Structured input tools let the clinician enter findings — diagnosis, examination, plan — and the AI generates a formatted UK clinic letter in under 30 seconds. No recording, no consent burden, no hallucinated content from misheard audio. The clinical facts come from you; the AI handles formatting.

There's an alternative approach to AI-assisted letter writing that avoids the recording problem entirely: structured input.

Instead of capturing everything that was said in the consultation, you enter just the clinical findings you want in the letter. A few bullet points — diagnosis, relevant examination findings, investigation results, management plan — and the AI generates the complete, formatted letter from those inputs.

This is how Docyment works.

I built Docyment because I was tired of writing letters at 10pm. The specific problem I wanted to solve wasn't "I don't know what to write." It was "I know exactly what I want in this letter, and it still takes days to appear."

With structured input, the AI formats what you told it — nothing more, nothing less — rather than interpreting what it thought it heard. There's no recording consent burden. And the output is a proper UK clinical letter — addressed to the GP, structured in the format UK consultants and GPs expect, ready to send the same day.

What to Look for in Clinical Letter Software

If you're evaluating options, here are the questions worth asking:

Does it produce UK clinical letters or American clinical notes? The format matters. A note optimised for Epic or Cerner is not a UK clinical letter. Check the output before committing.

Does it require recording the consultation? If yes, think carefully about your patient population and your GDPR obligations. For some practices this is acceptable; for others it isn't.

Who reviews the clinical content before it's sent? You should. Any tool that sends letters without clinician review is a risk. The AI should assist, not replace, clinical sign-off.

What happens with patient data? Where is it processed? Is it stored? Is it used to train AI models? Ask the vendor directly. For patient-identifiable clinical data, you need clear answers.

Does it integrate with your existing workflow or replace it? The best clinical letter software fits around how you already work — it doesn't demand that you rebuild your entire practice workflow to accommodate it.

Which Clinical Letter Approach Fits Your Practice in 2026?

If you know what you want in the letter, you don't need to record a consultation to get it. Structured input is faster, safer, and simpler for private practice consultants who write 10-20 outpatient letters per clinic.

Clinical letter software has improved significantly in the past three years. AI tools can genuinely reduce the documentation burden for UK consultants. The question is which approach — ambient recording or structured input — fits your practice.

My view, having used both approaches: if you know what you want in the letter, you don't need to record a consultation to get it. You just need a faster way to turn your clinical impression into a formatted letter.

That's what Docyment is built for.

Frequently Asked Questions

What is the best clinical letter software for UK private practice?

The best tool depends on your workflow. For consultants who know their findings and need formatted UK letters fast, structured AI tools like Docyment produce same-day letters in under 30 seconds. For high-volume NHS settings where note capture during the consultation is the priority, ambient AI scribes like Heidi or Tortus are better suited.

Is AI safe for writing clinical letters?

Yes, with the right guardrails. Structured input tools format facts you provide — the hallucination risk is negligible because the AI is not inferring clinical content from audio. Every letter should be reviewed and approved by the clinician before sending. The AI assists; you remain the author.

Do I need patient consent to use AI clinical letter software?

It depends on the type of tool. Ambient AI scribes that record the consultation require explicit patient consent under UK GDPR. Structured input tools — where you enter findings by typing, dictating, or scanning notes — carry no recording consent burden because there is nothing to record.

How much time does AI clinical letter software save?

A private practice consultant writing 12 letters per clinic typically spends 60-90 minutes on dictation. With structured AI, the same letters take 15-20 minutes including review. That is roughly an hour saved per clinic session, with letters sent the same day instead of 2-4 days later.

Clinical Letter Software for UK Consultants: What's Actually Worth Using