In late June, an oncologist treating a patient at a hospital in Boston opened an electronic chart and found something new sitting inside it: a decision support tool called Solavia, built jointly by NYU Langone Health and the Dana-Farber Cancer Institute, quietly surfacing the latest evidence and treatment pathways for that patient's specific tumour before the clinician had finished reading the notes. The tool has since gone from internal pilot to commercial product, with NYU Langone's technology arm now selling access to other health systems. Two of America's most prestigious cancer centres have effectively decided that their accumulated clinical judgment, built on their own patients and protocols, is now a product line.

There is no reason to anticipate any NHS involvement in Solavia. The interest for a British audience lies less in the product itself than in what it represents: a preview of how AI-assisted decision support in oncology is likely to be built and sold globally over the coming years, by a small number of well-capitalised academic health systems refining tools on their own patient populations before licensing them outward. That matters for a country whose cancer pathways are under sustained strain and whose government has just published a National Cancer Plan promising to meet every cancer waiting time standard by 2029, a target that will require closing gaps that have proven stubborn for a decade.

The scale of that gap is worth stating plainly. The headline 62-day standard, which requires 85 per cent of patients to begin treatment within two months of an urgent referral, has not been met since December 2015. Performance sat at 69.6 per cent in May this year, some fifteen points short. The 28-day faster diagnosis standard, meant to confirm or rule out cancer within a month of referral, is also missing its target. These figures are familiar to anyone working in NHS cancer services, but they matter here because standardising and speeding up clinical decision-making, precisely what tools like Solavia claim to do, is one of the few mechanisms with a plausible route to closing that gap without simply asking an already stretched workforce to move faster than it can.

That raises the harder question for NHS leaders and the life sciences sector watching this from a distance. If embedded, evidence-linked decision support becomes a meaningful factor in cancer outcomes, and there is a reasonable case that it will, the NHS faces a choice that has already played out once in a different part of its digital estate. It can import a tool built and refined on another health system's patients, protocols and legal assumptions, adapting a foreign clinical logic to British practice. Or it can build its own, more slowly, with less capital behind it and a workforce that has little spare capacity for product development on top of patient care. The country has just spent two years arguing over whether it was wise to let an American vendor build the platform that schedules its beds and tracks its waiting lists. Extending that same dependency into the substance of treatment recommendations, rather than the logistics around them, would raise the stakes considerably.

None of this means Solavia will ever reach a British hospital, and the connection here is a cautionary one rather than a live procurement decision. But the pattern it illustrates, academic medical centres becoming vendors of their own clinical intelligence, is unlikely to remain a uniquely American phenomenon, and the NHS's cancer pathways are exactly the kind of pressured, high-stakes territory where an imported solution will look tempting long before anyone has worked out what depending on it would cost. The lesson from Solavia has little to do with whether Britain ever adopts this particular tool. The NHS has very little time left to decide whether it wants to be a customer of this kind of intelligence or a producer of it, before that decision gets made for it by default.