A remote monitoring pilot ends on a Friday in March. The devices that let a community team watch a frail patient's oxygen levels from the office are collected into a box, because the grant that paid for them has run out. On Monday the nurses go back to telephone calls and visits prioritised by instinct. The tool worked. The evaluation says so. Nobody holds the budget line to keep it.
Anyone who has worked in community services will recognise that scene. It sits behind the argument made by Paulette Hamilton, a former district nurse and member of the Health and Social Care Committee, who says the NHS struggles to carry proven ideas beyond the site that first tried them. She is right, and the point deserves more scrutiny than the enthusiasm around artificial intelligence usually allows. The three shifts in the Government's 10 Year Health Plan all depend on technology travelling from trial to routine practice. The machinery for that journey remains poorly built.
Start with money. Pilots are typically paid for from short-term funds with horizons of a year or less. Trusts, by contrast, run annual budgets under tight financial controls, with waiting lists above seven million and deficits to manage. A chief executive asked to take on a recurring licence fee for a tool whose savings accrue in the ambulance service or in social care will usually decline, and the decision is rational. The benefit is real but sits in someone else's ledger. Hamilton's test of whether an innovation gives staff more time with patients is a sound clinical measure. It says nothing about who pays.
Procurement compounds the difficulty. Individual organisations still assess the same product separately, which burns months and drains the cash of small suppliers who cannot afford a hundred parallel sales cycles. The MHRA and NICE have improved the routes for establishing that a technology is safe and good value. Neither can compel adoption, which remains a local decision made by people with little spare capacity. The abolition of NHS England and the pressure on integrated care board running costs now remove some of the central and regional capability that might have brokered spread across many sites. Ministers will need to say who owns that task.
Digital foundations deserve the same candour Hamilton applies to them. Tools that draft clinical notes or letters are the nearest prospect for relief, and they perform well only where records are complete, systems communicate and the network holds up. Trusts differ widely in all three. Placing a sophisticated tool on fragmented records produces expensive disappointment, and the patients in Birmingham Erdington who repeat their histories to each new service will see little change.
Accountability is the least developed question. The principle that responsibility stays with human professionals is sensible, but a nurse who accepts or overrides an algorithmic recommendation needs more than the principle. She needs training in what the system can and cannot do, clarity on where liability falls, and evidence of how the tool performs for the population she serves. Hamilton's constituency has a younger burden of ill health than much of the country. A model trained mainly on data from elsewhere may serve it badly. Performance by ethnicity, age and deprivation should be a condition of adoption, published before purchase and not requested afterwards.
The implications differ by audience. Trust leaders need pooled or nationally held funding for tools whose benefits cross organisational boundaries. Policymakers need to decide whether adoption is a national duty or a local option, since the present arrangement treats it as the second while the plan assumes the first. Suppliers should expect evidence of equity and integration to matter as much as accuracy. Patients should be offered a non-digital route that is as good as the digital one, because a service that works only for the confident and well connected will widen the gaps it claims to close.
The Government will be judged by how many tools are in everyday use by the end of this Parliament, and pilots launched will count for little. Innovation reaches patients when someone is made responsible for paying for it, spreading it and answering for it. Until the plan names that someone, the box of devices will stay in the cupboard.
