My father is a specialist. He has been for most of my life. I grew up watching him come home from the hospital and sit at the kitchen table to type letters.
Long, considered letters to the GPs who had trusted him with their patients. The clinical content was thorough and precise, but the letters were never impersonal. They were the continuation of a professional relationship that, in some cases, had been running for decades.
I used to think this was just how he was. Careful. Particular. I didn't understand until much later that those letters were as much about the relationship as they were about the patient. The GP reading them wasn't just receiving a clinical update. They were being reminded that there was a colleague on the other end of this referral who respected them, who took this seriously, who could be trusted with the next patient too.
That is what a specialist's letter actually is. Not documentation. A professional handshake, repeated every time.
This is why we built Medow’s new Personalisation feature. And it's why we built it the way we did.
When we started building Medow, we heard a lot about efficiency. Specialists wanted to spend less time at the keyboard and more time with patients. That part we expected.
What we didn't fully anticipate was the subtlety of what happened when they started using AI. The clinical content was often good. The summaries were accurate. The structure was appropriate. But something kept coming up in feedback, in the calls we had with specialists who had tried AI and decided it was not for them.
"It doesn't sound like me."
That sentence, or some version of it, came up again and again. Not "it gets things wrong." Not "it's too slow." The friction was something harder to define and harder to fix.
One psychiatrist described it as reading someone else's handwriting. The words were right. The meaning was there. But the voice was off in a way she couldn't articulate beyond knowing it immediately.
A gastroenterologist told us he spent longer editing his letters than he had spent dictating them. Not because anything was incorrect. Because every time he read a draft, he found another phrase he would never use, another opening line that sounded like AI had read a lot of gastroenterology letters and produced a statistically likely one.
We listened. We sat with that feedback for a long time. We consulted with our advisory board of specialists. Because fixing it wasn't a matter of adding a toggle or adjusting a template. It required rethinking what personalisation actually means in a clinical context.
Every clinician who has been writing for more than a few years has developed what linguists call an idiolect: a distinct fingerprint of vocabulary, rhythm, and structure that is entirely their own.
The words you reach for first. The way you order your clinical reasoning. Whether you lead with the diagnosis or the history. How you address a colleague. How do you close a letter when the news is complicated.
These are not style choices. They are professional identity. They have been shaped by your training, your specialty, your mentors, your patients, and your colleagues recognise it. Patients come to trust it. The GPs who refer to you know, from the first paragraph, that it is you.
and
For specialists, this matters in a way that goes beyond efficiency. A referral letter is not the same as a clinical note. Where a GP's notes are primarily functional records, accurate and internal, a specialist's letter is a professional communication between colleagues. It carries clinical reasoning and recommendations to a colleague who has placed their trust in that referral relationship. The tone, structure, and language of that letter directly reflect the specialist's professional standing. Referring GPs read these letters repeatedly over the course of long careers. They know, from the first paragraph, whether it sounds right. When it doesn't, when the voice is slightly off or the phrasing a little generic, it registers. Not as an error. As a subtle inconsistency that erodes the sense of a known and trusted colleague on the other end.
When AI generates output that doesn't carry your fingerprint, you feel it immediately. And you fix it. Every time. Because you cannot send something that doesn't sound like you. Not to the GP who has referred patients to you for a decade. Not under your name.
Medow is a specialist-grade AI, built from the ground up for each specialty, with practising clinicians involved in its development at every stage. That means the model already understands what standard psychiatry correspondence looks like, what is typical in cardiology, and what the structural conventions of your specialty are. That baseline is not a minor technical detail. It is what makes true personalisation possible.
We spent a long time getting that foundation right. We didn't want to offer personalisation until we could offer it properly. Until we could look a specialist in the eye and say: this will actually sound like you.
The timing of this release came directly from customers. Specialists who had tried AI and stopped. Specialists who stayed but told us the efficiency gain was real but incomplete because the voice was still theirs to fix. That feedback shaped this feature more than any internal roadmap did.
Three letters. That is all it takes.
Upload three examples of your existing clinical letters. Within minutes, Medow has a working profile of how you write: the vocabulary you prefer, the structure you use, the tonal register you have spent a career developing.
From that point on, every report Medow generates carries your idiolect, not a version of what an AI thinks a clinician in your specialty might write. Your word choices. Your structure. The way you address a colleague. The way you sign off.
The manual editing doesn't disappear entirely, and we wouldn't claim otherwise. But the gap between what the AI produces and what you would have written closes substantially. The letters that go out under your name start to sound, again, unmistakably like you.
And the vision extends beyond the referral letter. Consult notes that a doctor checks when a patient returns. Op reports for procedures. Patient letters are written so that patients can understand them. Every report type a specialist produces can carry the same voice. The letter is the crown jewel, but personalisation applies to the full suite.
I think about my father at the kitchen table. About the care he took. About the relationships those letters represented.
He wasn't writing to discharge an obligation. He was maintaining something, nurturing something, that had real value for his patients and for the colleagues who trusted him with them. The letters were the medium. The relationship was the point.
Specialists build those relationships over careers. The trust embedded in a referral relationship is not given easily. It is earned through every interaction, every letter, every consult note, every op report, and every patient letter that arrives and sounds exactly as it should.
AI should not threaten that. It should protect it.
That is what we have built. Not a tool that sounds like a clinician. A tool that sounds like you.
