NHS AI Faces Resistance From Manchester to the GP Phone Line

Resistance to artificial intelligence in Britain’s health system is no longer confined to abstract arguments about privacy or innovation. It is now showing up in two very different places: in a major regional challenge to a national NHS data program led by Palantir, and in the everyday frustrations of patients who say an AI receptionist cannot understand them when they call their doctor.

In Greater Manchester, the country’s second-largest urban region has renewed its refusal to formally adopt the NHS Federated Data Platform, the £330 million, seven-year system awarded in 2023 to a Palantir-led consortium. In Rotherham, meanwhile, patients have complained that an AI telephone receptionist known as Emma has struggled with broad Yorkshire accents and some speech impairments, leaving some callers so frustrated that they hung up or went to their GP surgery in person.

The disputes are separate, but together they suggest a broadening skepticism over how artificial intelligence is being introduced into the National Health Service: from top-down procurement of vast data infrastructure to frontline tools meant to make basic access to care easier.

A regional holdout

Greater Manchester has become the lone English integrated care board to resist formal adoption of the Federated Data Platform, arguing that its own local data systems are better suited to the region’s needs. The stance matters not only because of the size and influence of the area, but because NHS England has made clear that it wants the platform to become a core part of how providers and regional bodies plan services and coordinate care.

Guidance updated this spring said the medium-term planning framework expects all providers and integrated care boards to be onboarded and using the platform’s core products by 2028-29. The platform is designed to help NHS organizations combine and analyze data across hospital backlogs, bed management, waiting lists and broader service planning.

But the promise of a single national platform has collided with long-running anxieties over centralization, procurement and trust. Palantir, the American data company best known for its work with governments and security agencies, has faced years of criticism from campaigners and some clinicians who question whether such a large role in the NHS should be handed to a company with that history, even though NHS officials have insisted the platform does not alter existing rules on data control or patient confidentiality.

Pressure has intensified this year. The BMJ reported in July that NHS chief executive Jim Mackey had supported an independent review of the platform, while ministers have acknowledged that the contract contains break clauses. One key decision point is expected around February 2027, leaving open the possibility that the government could alter course if the review is damaging or if take-up remains uneven.

Questions have also been raised about whether the platform is delivering the benefits promised. BMJ reporting this year examined the evidence used to support some of those claims and said roughly a third of trusts with access to the apps appeared not to have used them over the previous year. For critics, that has strengthened the case that local NHS organizations should not be pushed into a one-size-fits-all model if existing systems already work.

Greater Manchester’s resistance, then, is not simply symbolic. It tests whether ministers can make the case that a centrally procured data platform is indispensable, even when a prominent region insists it can do the job itself.

Friction on the front line

If the fight over Palantir is about strategy and governance, the complaints in Rotherham are about something more immediate: whether patients can get through to their doctor.

Healthwatch Rotherham, a local health and social care watchdog, said patients had reported problems with Emma, an AI receptionist introduced at a number of local GP practices. According to the watchdog, the system struggled with broad local accents — the “twangs,” as one account put it — as well as with some speech impairments. Some patients reportedly abandoned calls out of frustration; others turned up at surgeries in person instead.

For GP practices under immense pressure, AI reception tools have been pitched as a way to manage demand, shorten waits on the phone and direct patients more efficiently. Vendors say such systems can work across multiple languages and free up staff for more complex tasks. But the Rotherham complaints point to a recurring problem in AI deployment: a system that functions well in theory or in controlled demonstrations can fail in ordinary use, especially when confronted with regional speech, disability-related communication differences or patients who are already anxious, unwell or digitally excluded.

That matters in primary care, where contact itself is the gateway to treatment. If a patient cannot make an appointment request, describe symptoms or navigate a call system, the technology does not merely inconvenience them; it can become a barrier to care.

The risks are particularly acute for older people, those with disabilities, and patients who may not be comfortable with automated systems to begin with. In a health service already struggling with access and trust, even small failures at the reception stage can carry outsized consequences.

A wider test for the NHS

Artificial intelligence has been promoted across the NHS as both a practical necessity and a route to modernization in an overstretched system. Officials and technology suppliers argue that better use of data can improve planning and resource allocation, while automation can ease administrative bottlenecks that consume staff time.

Yet the current backlash underscores that the barriers to adoption are not just technical. They are political, social and deeply human.

At the national level, the Palantir dispute raises the question of whether the NHS can build legitimacy for a data program of this scale if some local systems believe they are being asked to replace functioning arrangements with a platform they neither chose nor clearly need. At the practice level, the Rotherham case highlights a different kind of legitimacy problem: whether patients will accept automation that appears to make basic interactions harder rather than easier.

Neither debate is settled. It remains unclear whether the government will stay the course with Palantir through the full term of the contract or invoke a break clause, and whether Greater Manchester will remain an isolated outlier or become a model for other regions seeking more autonomy. It is also not yet known how widely tools like Emma are being used, or whether regulators and local NHS leaders will tighten expectations around accessibility, accent recognition and a reliable human fallback.

What is clearer is that the argument over AI in British health care has moved beyond aspiration. It is now being judged on evidence of benefit, on local control, and on whether real patients can use the systems placed in front of them.

Sources

Further reading and reporting used to add context:

AI News