At a GP practice in Sussex, a trial technology has been quietly proving its worth. Patients tell an app what is wrong with them, and the app tells them where to go: pharmacy, GP, A&E, or nowhere at all. The result was a 29 percent drop in patients queuing on the phone, the kind of number that makes for an easy press release. On Monday it became the centrepiece of NHS England's announcement of how it will spend £10 billion on technology and data over the next three years, a package the service says will eventually return £41 billion in benefits and deliver half the commitments in the government's own 10 Year Health Plan.

It is, on its face, a serious and well-evidenced piece of policy. It is also being announced into a department, and a government, that currently cannot say with any confidence who will be running either of them by the fourth or fifth year of the rollout.

James Murray became health secretary in May, one of several ministers parachuted in after Wes Streeting resigned amid Labour's post-election turmoil. Streeting had built the 10 Year Health Plan that this technology programme is meant to help deliver. Murray inherited the plan without having authored it, a common enough occurrence in Whitehall but not usually while the prime minister who appointed him is simultaneously serving out his own resignation. Keir Starmer announced in June that he would step down, and remains in post only until Labour completes a leadership contest that opens for nominations this Thursday. The overwhelming favourite to succeed him is Andy Burnham.

None of this makes the AI triage tool a bad idea. The clinical logic behind it, and behind the parallel expansion of ambient voice technology that listens to consultations and drafts the notes, is sound enough that trusts have been asking for exactly this kind of investment for years. Great Ormond Street's study on ambient documentation, showing clinicians gaining back close to a quarter of their patient-facing time, is a genuinely persuasive piece of evidence, not a marketing figure dressed up as one.

What is harder to square is the confidence of the timeline against the instability of the mandate. NHS App coverage for all users is promised by April 2028. Getting there requires sustained ministerial attention, sustained capital allocation, and a workforce willing to keep integrating new systems into daily practice through at least one and probably two further changes of health secretary, and quite possibly a change of prime minister within weeks. Digital programmes in the NHS have a long history of losing momentum precisely at these junctures, not because the technology fails but because the political sponsorship that protected it moves on.

There is a second layer to this, one NHS England did not choose to foreground in its own announcement. The MHRA's recent National Commission into the Regulation of AI in Healthcare found that almost three quarters of respondents did not believe the current regulatory framework was adequate to guarantee safety and performance. Nearly two thirds raised concerns about data governance, and a similar proportion about post-market surveillance once tools are in live use. These are not fringe objections. They originate from individuals operating within the system through which this rollout will take place, and they indicate a disparity in confidence between what is implemented and what the regulatory framework can presently substantiate.

Burnham's own record on the NHS, built through years of devolution work in Greater Manchester, points to a different instinct on digital delivery than Streeting's more centralised approach, and a change of leader tends to bring a change of priority, whatever officials insist about continuity. Jim Mackey's enthusiasm for the programme is genuine, but enthusiasm from NHS England's chief executive cannot substitute for a stable ministerial hand holding the policy in place.

The triage tool works. The question worth asking is not whether the technology can perform, but whether the government announcing it will still exist in a form capable of finishing what it started.