The 8 a.m. rush for GP appointments at Wealden Ridge Medical Partnership, a rural clinic with four locations in Sussex, used to result in a phone queue that both patients and staff feared. Since the practice began trialling an AI triage tool built into the NHS App, that queue has fallen by 29 percent. Patients describe symptoms through the app, the system asks follow-up questions and requests photographs where useful, and a clinician reviews the resulting summary before deciding what happens next. Dr Ragu Rajan, the GP who ran the trial, says the technology has not replaced clinical judgement so much as freed up time to exercise it properly.

That is a genuinely encouraging result from a single site serving 23,000 patients. NHS England has now decided it is encouraging enough to build a national programme on. Over the next twelve months the tool will reach more than 200,000 patients in London, and by April 2028 it is due to be available to every NHS App user in England, folded into a wider technology overhaul worth £10 billion over three years that also includes ambient voice notetaking for clinical consultations. NHS England's chief executive, Sir Jim Mackey, has described the changes as transformative for how patients reach the right service for the first time.

The gap between those two claims deserves more scrutiny than it has received. A rural practice with an engaged GP and a manageable patient list is not a representative test bed for how an urban population, with wider variation in digital access, language, and trust in institutions, will use the same tool. Twenty-nine percent at Wealden Ridge is a real number, but it describes one context, not a system. Extrapolating from it to a national rollout with a fixed delivery date treats a promising pilot as though it were already proven at scale.

The regulatory picture makes this harder to wave through. The MHRA's own National Commission into the Regulation of AI in Healthcare put these questions to the sector directly, and the results were not reassuring for anyone hoping the framework is ready. Nearly three quarters of respondents doubted the current system was sufficient to guarantee safety and performance standards. Well over half thought existing requirements for clinical evidence fell short, and a similar proportion said data governance and post-market surveillance both needed further work. This is not scepticism imported from outside the health service. It is the regulator's own commissioned findings, published alongside a rollout timetable that assumes those gaps will be closed in time.

There is also a less discussed problem, which is who will actually be accountable for any of this by 2028. The rollout was announced under James Murray, who became health secretary in May after Wes Streeting's abrupt departure and lasted barely ten weeks in the role. This week Andy Burnham, newly installed as prime minister, replaced Murray with Yvette Cooper, returning to the health brief she last held under Tony Blair. Cooper inherits a technological program whose completion date is well beyond the current political upheaval and potentially even beyond her own term. NHS technology commitments of this scale routinely outlive the ministers who announce them, which makes it easy for accountability to blur into something closer to institutional memory than personal responsibility.

None of this means the triage tool is a bad idea. Reducing pressure on GP phone lines while keeping clinicians in charge of decisions is a reasonable goal, and Wealden Ridge suggests it can work somewhere. But a single pilot, an unresolved regulatory gap acknowledged by the regulator itself, and a health secretary already three names deep since the policy was announced are not the ingredients of a settled programme. They are the ingredients of one that needs harder questions asked of it before 200,000 Londoners become the next test site.