Last updated: April 2026

I have been a doctor for twenty-six years and a Consultant Urologist at Kettering General Hospital since 2022. But the problem I am about to describe started long before Kettering. Writing clinic letters was something I accepted the way you accept weather. It happened after every clinic. It took however long it took. You got on with it.

A few years ago I stopped accepting it. Not because I had a technological epiphany. Because I sat down one Thursday evening with a stack of fourteen letters still to dictate, realised I had been doing this exact thing every week for over two decades, and thought: this cannot be the best we can do.

What followed was a conversation with my cousin Sherif — a software developer — that turned into a product. That product is called Docyment. This is the story of why it exists, what it does, and what it does not do.

What Problem Does a Consultant Urologist Have with Clinic Letters?

Every time a UK consultant sees a patient in clinic, a formal clinical letter must be produced. Not a note in a file. A letter addressed to the patient's GP, structured with the right headings — history, examination, investigations, clinical impression, management plan, follow-up — in the right tone, with the right detail. It is a medicolegal document that may be read by an insurer, a solicitor, or another consultant years from now.

There is a version of this story where a founder says "clinical documentation is broken" and gestures broadly at the state of healthcare administration. That is not this story.

My problem was narrow and concrete. I see twelve to fifteen patients in a Friday afternoon urology clinic. That means twelve to fifteen of these letters, each one taking five to seven minutes if I dictate them efficiently. That is an hour and a half of work that begins after the clinical work is already finished.

Multiply that across a week of clinics — NHS and private — and the number climbs to forty, fifty, sometimes sixty letters. This is not an inconvenience. It is a second job.

And it gets worse the longer you defer it. A letter dictated immediately after the patient leaves takes five minutes. The same letter dictated at 9pm, reconstructed from notes and a fading memory of the conversation, takes eight or ten. By then you are second-guessing the exact PSA value, whether the MRI showed a PI-RADS 3 or a PI-RADS 4, whether you told the patient six weeks or eight weeks for the follow-up. This is what I call documentation debt — and like financial debt, it compounds.

The clinical content degrades. The time expands. The evening disappears.

Why Did Existing Tools Not Solve the Clinic Letter Problem?

Before building anything, I tried every available solution. Dragon Medical dictation, outsourced transcription services, and ambient AI scribes all address parts of the problem — but none solve it for a UK consultant producing formal clinical correspondence.

Dragon Medical dictation. I used Dragon for years. It is good at turning speech into text. But it does not write a letter. It gives you a transcript of what you said, and you still have to format it, structure it, address it, add the clinical headings, and review the whole thing. The letter still takes five minutes. You are just speaking the five minutes instead of typing them.

Outsourced dictation services. I tried sending audio files to a transcription service. The turnaround was 24 to 48 hours. The quality was inconsistent. The typist did not know my patients, my style, or the difference between a urology follow-up letter and a new patient letter. I spent as much time correcting their output as I would have spent writing the letter myself.

Ambient AI scribes. More recently, ambient tools — Heidi Health, Tortus, Nuance DAX — have entered the market. They record the consultation, transcribe it, and generate a clinical note. For GPs writing into EMIS or SystmOne, where the output is a record entry rather than a formal letter, they work well.

But my output is not a record entry. It is a structured letter to a named GP, with specific sections in a specific order, in a formal register. Ambient tools generate notes. I need letters. The gap between those two things is where the editing time lives. You get a summary, you rewrite it into a letter, you add the structure, you adjust the tone — and you are back to spending five minutes per patient, just with a different starting point.

I should be honest: ambient recording is genuinely better suited for certain workflows. If you are a GP seeing forty patients a day and your output is a clinical note that goes straight into the patient record, ambient makes sense. If you are a secondary care consultant producing formal correspondence, the fit is less obvious. For a full comparison of both approaches, see AI medical letters UK: what consultants need to know.

An independent validation study of Tortus AI found a 1.47% hallucination rate in generated content, with 44% of those errors rated clinically significant. In a clinic note reviewed within the EHR, that may be acceptable with careful checking. In a medicolegal letter addressed to a GP and potentially forwarded to the patient, a single fabricated finding is a different kind of risk.

I am not criticising these tools. They solve a real problem for the right user. They just did not solve my problem.

How Did Docyment Start?

The decision to build came from a single question. I described the workflow I wanted to my cousin Sherif: I see a patient, I capture the key clinical findings, and something produces the finished letter — addressed, formatted, structured — within a minute. Not a draft I have to reshape. A letter I can review and send.

Sherif asked me: "Can you describe the letter format precisely enough that a machine could write it?"

The answer was yes. UK consultant clinic letters follow a predictable structure. The headings are consistent. The tone is consistent. The level of detail expected by a GP is consistent. What varies is the clinical content — the findings, the results, the plan. If I could give the system the clinical content and the template, the system could produce the letter.

The first version was rough. It worked for urology outpatient letters only. The template was rigid. The output needed heavy editing. But the core insight held: if you separate the clinical content from the letter structure, and you have a system that understands both, you can generate a finished letter from a handful of clinical inputs.

We iterated. We added more templates — post-operative, follow-up, new patient, insurer reports. We added Sherif Darwish, a Consultant Psychiatrist in Leicester, who brought a completely different letter format and validated that the approach worked across specialties.

What Does Docyment Actually Do?

Docyment is a template-based AI clinical letter generator with four input modes. You select a patient, choose a template, enter the clinical content in whichever way suits you, and receive a finished UK clinic letter in under 30 seconds. Every letter is reviewed by the clinician before it goes anywhere.

The workflow today is this. You open Docyment, select a patient, and choose a template. I have templates for urology outpatient, post-operative follow-up, LUTS assessment, haematuria investigation — each one structured the way I actually write those letters.

Then you give it the clinical content. You have four ways to do this:

Dictate — you speak your findings directly into the app. Not a full letter dictation. Just the clinical content. "PSA 8.2, up from 4.1 last year. MRI shows PI-RADS 4 lesion right peripheral zone. Discussed findings and recommended template biopsy. Patient consents. Listed."

Record — if you prefer, Docyment records the consultation itself. It extracts the clinical content from the recording and applies it to the letter template.

Write — you type the clinical findings. Four or five bullet points. This is what I use most often in my own practice. I can type the key details in thirty seconds between patients.

Scan — you photograph your handwritten clinic notes and upload them. Docyment reads the handwriting, extracts the clinical content, and generates the letter.

You press Process Letter. In under thirty seconds, you have a complete, properly formatted consultant letter. Your name and credentials. The GP's name and address. The clinical headings in the right order. Ready to review and send.

Every letter is reviewed by the clinician before it goes anywhere. You are always the author of record. The AI assists with the formatting and structure; the clinical judgement is yours.

What Does Docyment Not Do?

I want to be direct about the limitations, because clinicians can spot a sales pitch from across a room.

No EHR integration. Docyment does not write into EMIS, SystmOne, or Cerner. It generates a letter that you review, approve, and then send or upload through your existing channels. For private practice — where most consultants email letters or print them on headed paper — this is not a limitation. For NHS trusts with mandatory EPR workflows, it is. EHR integration is on the roadmap, but I will not claim it exists today.

No clinical judgement replacement. It does not diagnose. It does not suggest management plans. It does not infer findings you did not provide. If you tell it the PSA is 8.2, it writes 8.2. If you do not mention the MRI, the letter will not contain MRI findings. The clinical content comes from you.

Not ambient-only. Record is one of four input modes. If ambient recording suits your workflow, use it. If you would rather type three bullet points between patients, do that instead.

No AI training on patient data. Your clinical data is yours. It is encrypted end-to-end. It is never used to train or improve AI models. GDPR compliant. This was non-negotiable from the start.

Why Does a Clinician-Founder Build Differently?

There is a difference between a product built by someone who researched the problem and a product built by someone who lives it every Friday afternoon. A clinician-founder knows the clinical workflow from the inside — not from user research sessions, but from twenty-six years of writing the letters.

I know what a urology follow-up letter looks like because I have written thousands of them. I know that the GP wants the PSA trend, not just the latest value. I know that "plan: continue watchful waiting" is meaningless without specifying the interval and the trigger for escalation. I know that an insurer letter has different headings from an NHS letter, and that a post-operative letter to a GP needs the operation name, the findings, and the discharge plan in that order.

When Sherif Darwish joined as our third co-founder, he brought the same depth from psychiatry. His letters have a completely different structure — longer histories, different clinical terminology, different expectations from the receiving GP. The system had to accommodate both. It does, because the people building it write these letters themselves.

Docyment is built by two practising consultants and a developer. Not a product team that interviewed some doctors. Three people who felt the problem, designed the solution, and use it in their own clinics.

Where Is Docyment Now?

Docyment is in early access. The product is live at app.docyment.com. You can sign up today and start writing letters.

The free tier gives you twenty letters per month, one hundred patients, and ten templates. No credit card required. No setup call. No trial that expires. Twenty letters is enough to run one or two clinics and decide whether it works for you.

For consultants who want to use it as their primary letter tool, the early-adopter Pro plan is £20 per month — unlimited letters, unlimited patients, unlimited templates, letter export. That price locks in permanently. When we move to standard pricing, early adopters keep their rate.

I use Docyment for every clinic I run — NHS and private. I finish my letters before the last patient has left the building. Most of those letters arrive in the GP's inbox the same day.

Frequently Asked Questions

Is Docyment an AI scribe?

Not exactly. An AI scribe records the consultation and generates a note from the transcript. Docyment is an AI clinical letter generator with four input modes: Dictate, Record, Write, and Scan. Recording is one option, not the only one. The output is a formatted UK clinic letter, not a clinical note.

Can I use Docyment for private practice and NHS clinics?

Yes. Docyment generates standard UK clinical letters that work for both private practice and NHS correspondence. The same templates, the same workflow. I use it for both in my own practice at Kettering General Hospital.

How is Docyment different from Dragon Medical dictation?

Dragon turns speech into text — you still have to format, structure, and address the letter yourself. Docyment takes your clinical findings (by any of four input methods) and produces a complete, formatted clinic letter with proper headings, GP addressing, and clinical register. The difference is between a transcript and a finished letter.

What specialties does Docyment support?

Docyment currently supports urology and psychiatry with validated templates. The template system is designed to accommodate any specialty — you can create custom templates matching your letter format. We are expanding specialty coverage based on early adopter feedback.

Is patient data safe with Docyment?

Patient data is encrypted end-to-end and never used to train or improve AI models. Docyment is GDPR compliant. The clinician reviews and approves every letter before it is sent. Your clinical data remains yours.

Why I Built an AI Clinical Letter Generator: A UK Consultant's Story