NHS trusts face huge admin pressure and a clear policy steer toward AI — but adoption runs through specific NHS gates. Here’s how trusts can adopt AI sensibly in 2026. (dgm is an independent integration partner that implements osFoundry; we are not osFoundry, and clinical responsibility stays with the trust.)

The policy backdrop

The Fit for the Future: 10 Year Health Plan for England (July 2025) names data, AI, genomics, wearables and robotics as priority transformative technologies. So the direction is set — the question is how to adopt AI within NHS rules.

High-value use cases

The strongest starting points are non-clinical:

  • AI-enabled ambient scribing — NHS England published guidance on ambient scribing in March 2026, setting IG and consent expectations;
  • correspondence and referral triage;
  • knowledge retrieval across clinical guidance; and
  • back-office automation.

These augment staff and cut admin burden — clinical judgement and accountability stay human.

The gates: DTAC, IG, DSPT

This is what’s distinctive about NHS AI. Adoption is gated by:

  • DTAC — the national baseline covering clinical safety, data protection, technical security, interoperability, and usability/accessibility (refreshed form transitioning fully by April 2026);
  • NHS Information Governance; and
  • the Data Security and Protection Toolkit (DSPT).

A tool can be funded and still fail to deploy if it doesn’t meet these — so build for them from the start.

The medical-device line

AI used for a medical purpose (diagnosis, treatment decisions) is a regulated medical device under the MHRA. Non-clinical use cases avoid this; clinical AI brings conformity and post-market obligations. Start non-clinical.

Where osFoundry and dgm fit

dgm implements NHS-grade AI on osFoundry with the controls trusts need: data control (self-host in your own cloud or an EU region — it publishes US/EU/JP regions, not a UK one), audit logging, and human oversight designed around DTAC, IG and DSPT. Patient data is special-category, so we design for IG and minimise exposure. osFoundry’s data-residency regions are US/EU/JP, so UK patient data goes to a self-hosted or EU deployment.

dgm is an independent integration partner with zero integrations so far — no NHS case studies to claim, just an honest approach. Clinical responsibility and medical-device compliance stay with the trust. To scope a compliant NHS AI project, book a consultation with dgm. Not clinical or regulatory advice.