About
One product. One job.
Radiology Assist exists to do a single thing exceptionally well: turn an MSK radiologist's dictated findings into a structured draft MRI report, in their own reporting style.
Radiology reporting has a strange shape: the clinical judgement takes minutes — the structure, the phrasing and the formatting take longer.
General-purpose AI scribes try to solve this for every clinician at once, and end up shallow everywhere. We went the other way. Radiology Assist is built in the UK with Dr Julian Chakraverty (MBChB BSc(Hons) PGDipSEM MRCS FRCR), a consultant MSK radiologist working in private practice, around one workflow: dictate, structure, review, export.
It learns each radiologist's reporting voice from their own previous reports, so the draft that comes back already sounds like them. And it stays in its lane: the AI never interprets images and never generates findings — it structures what the radiologist dictated, flags what looks incomplete, and hands control back.
What we believe
Three principles, no slogans
Narrow and deep
We are not a platform for every clinician. We build for MSK MRI reporting and go deep — region-specific structure, real reporting workflows, and quality checks that understand the anatomy.
Your style is the product
A report that doesn't sound like you isn't finished. Style learning isn't a feature bolted on at the end — it's the first thing the product does, before it generates a single word.
The radiologist stays in control
The AI structures; you decide. Every output is a draft, every draft requires your review, and nothing carries your name without your sign-off. We say "AI-assisted" and mean it.
Early access
Talk to the people building it
We're in early access and work directly with the radiologists using the product. Questions, feedback, or interest for your practice — we read everything.