The AI-Ready GP System: What Open Standards Mean for How You'll Learn Medicine

UK general practice is moving, slowly, from free-text notes held in one supplier's system to records that travel. The driving force is a set of open standards — published rules for how clinical information is structured, exchanged, and read by software other than the one that wrote it.

You do not need the spec name to understand the shape of the change. The change is this: a GP record is becoming something a different piece of software can read, query, and reason about, not just display.

For trainees, that is the part worth sitting with.

The quiet change happening in the GP record

For most of the last two decades, the GP record has been a free-text document with coded overlays — a problem list here, a prescription there, and a long narrative note in between. The narrative was the clinical truth of the consultation. The codes were what the system could count.

The standards now being adopted across the UK push in the opposite direction. They define how clinical information is represented in a structured form that any compliant system can read. The narrative becomes one element among many, not the spine.

The motivation is not fashion. It is the boring, important problem that the NHS has been trying to solve for a decade: when a patient moves from one part of the system to another, their record should travel with them, in a form the receiving clinician can trust.

Why this matters for how you learn medicine

When the record is a free-text blob, the doctor's job is partly to remember it. When the record is structured — coded problems, coded medications, coded observations against a date — the doctor's job shifts. The system can flag the duplicate prescribing, the missing review, the abnormal trend. The doctor does not have to hold all of that in their head.

The skill that does not move is reasoning. Knowing which coded problem to choose. Knowing when a coded entry does not capture what is happening. Knowing when to override the structured prompt and write a free-text note anyway. Knowing how to talk to the patient in front of you while the structured record builds itself in the background.

A small set of new suppliers are building AI-native GP systems on top of this shift. Their public pitch is that ambient scribing and structured data entry reduce the time a GP spends typing and increase the time they spend with the patient. Some of those systems are deployed in live UK practices. None of this changes the day-to-day skills a trainee is examined on. It changes the surface they are performed on.

What stays at the centre of the assessment

The GP training assessment, in both the recorded consultations and the applied knowledge test, scores two things above all others:

• Clinical reasoning. The ability to take a presenting complaint, form a differential, narrow it with focused history and examination, and arrive at a safe management plan.
• Communication. The ability to explain, to check understanding, to share decisions, and to record what was discussed.

Open record standards do not replace either. A system that listens to the consultation and writes the note can free up the doctor's attention. It does not generate the differential. It does not explain the diagnosis to a worried parent. It does not check that the patient understood the safety-netting advice. Those remain human tasks and remain the things examiners score.

What this changes for trainees in practice

Three practical shifts are worth naming:

• Note-writing becomes lighter. You will spend less time typing and more time looking at the patient. Trainees who learnt to write long free-text notes will need to learn to write short, structured entries that the system can read.
• Coding becomes a clinical skill. Picking the right code from a long list is part of the consultation, not an afterthought done at the end of the day. The choice of code carries information to the next clinician who reads the record.
• Audit and recall become automated, not abandoned. The system will pull the list of patients on a particular medication who have not had the required review. Your job shifts from running the search yourself to acting on the result.

None of these are assessed by examination today. All of them shape the day-to-day work a GP trainee is judged on in their supervisor review, and the supervisor review feeds back into the training assessment.

What to practise now

The skills that the open-record shift does not touch are the skills to drill.

• Run more cases. Not fewer. The structured record will summarise what you said. It will not tell you whether your reasoning was sound. Practice cases with examiner feedback do that — they score the reasoning, not the typing.
• Practise explaining. The patient still needs to leave the room understanding what you said. Record yourself, listen back, count the times you checked understanding. Structured records do not score this; examiners do.
• Practise writing in the structured style. Short sentences. Coded problem lists. A free-text line for the human detail. The system will read the structured part; the next clinician will read both.
• Treat coding as a habit, not an admin task. The right code in the right place is part of the consultation.

A note on the technology itself

The new GP systems being built on open standards are early-stage. Some are in pilot. Some are in production in a small number of practices. None has displaced the existing installed base, and none will this training year. What is certain is that the direction of travel — more structure, more exchange, more software reading the record — has been set by the published standards and is not going back.

That direction is good for patients. It is good for safety. It does not change what a good GP trainee looks like in an examination, and that is the part worth holding on to as the surface changes.

What structured means in practice

Structured does not mean rigid. A structured entry has three things: a coded problem, a coded medication or intervention tied to it, and a date. A free-text line sits alongside for the human detail that does not fit a code. The combination is what the next clinician reads, and what a different piece of software can also read.

A worked example. A patient with type 2 diabetes attends for review. The coded problem is the diabetes code. The coded observation is the HbA1c value. The coded medication is the metformin prescription. The free-text line is the bit about the patient's mother being diagnosed last month and the patient being more worried than the last time. Each part is doing a different job. The system reads the first three. The next clinician reads all four.

The skill trainees will be marked on is not which code they pick — that becomes muscle memory. It is whether they pick one at all, and whether the free-text line says what the codes cannot. That mix is what an examiner or supervisor will look for in a recorded consultation, and it is the same mix whether the record is written by hand or drafted by an AI.

What this post is not

This is not a product review. It is not a forecast of which supplier will win. It is not a defence of any specific AI scribe. It is a description of a trend that any trainee who reads the published standards can see for themselves, and a note on which of their existing skills the trend does not displace.

Practise the cases that score reasoning and communication: Start a free station.

← Back to all posts