GP practices are buried in admin — exactly where AI can help, if done within NHS data rules. Here’s how UK practices can adopt AI in 2026. (dgm implements osFoundry as an independent partner; clinical responsibility stays with the practice.)

Where AI helps GP practices

The high-value, lower-risk use cases are administrative:

  • documentation — ambient scribing, letter and report drafting;
  • correspondence and referral triage — organising and prioritising incoming items;
  • patient-query handling — routine, non-clinical questions; and
  • back-office — recalls, admin workflows.

All with clinician review — AI cuts admin so clinicians spend more time on care, while clinical decisions stay human.

The rules

GP practices are largely independent contractors under NHS contract, carrying data-controller responsibilities. The relevant gates:

The clinical line

AI that makes clinical decisions may be a regulated medical device (MHRA). The faster, lower-risk path is AI that handles admin and surfaces information for clinicians — not autonomous clinical triage.

The common risk

The biggest practical risk is staff using consumer AI tools with patient data — a data-protection problem. An approved, governed setup that keeps patient data in your control prevents it.

Where osFoundry and dgm fit

dgm implements IG-aware AI on osFoundry: data control (self-host or an EU region — it publishes US/EU/JP regions, not a UK one), audit logging, human oversight, and bring-your-own-key so patient data isn’t sent to consumer tools. Focused on admin use cases that deliver quick relief without entering medical-device territory.

dgm is an independent integration partner with zero integrations so far — no GP case studies to claim. Clinical responsibility stays with the practice. To scope an IG-compliant AI project, book a consultation with dgm. Not clinical or regulatory advice.