Medicine has become very good at producing clever things. AI can read images, genomics can divide diseases into ever finer categories, and diagnostics can tell us things that would have seemed improbable when I was a young doctor in Albania’s capital, Tirana. But we are less good at making sure this cleverness changes what happens to the patient.
I have spent much of my career thinking about how technology could change medicine. A year spent running my home country’s health system has made me think rather more about what medicine has to change if technology is actually to matter. Perhaps that is the real innovation.
An algorithm can perform beautifully in a study but make almost no difference on a Monday morning in a busy hospital. If AI reads a scan faster but the patient still waits weeks for the next step, what exactly have we transformed?
Instead, I would start somewhere unfashionably simple: with the clinical problem, not the technology. That distinction matters. Not everything can be a priority, nor is there unlimited money or workforce. A health service with 20 priorities risks having no real priorities at all.
In Albania, I chose a small number of connected areas: oncology, diagnostics, digital health and AI. The lesson was not that these were the right priorities for every country. It was that choosing mattered. Healthcare is often much more comfortable adding than choosing: another initiative, another platform, another target, another programme. Leadership sometimes means deciding what not to do.
Cancer care made another problem impossible to ignore. We talk endlessly about patient-centred medicine, then ask patients to carry information between hospitals, repeat their histories, chase appointments and navigate departments that do not always speak to one another. In complex cancer care, the patient and family can become the unofficial coordinators of the disease. That is a peculiar definition of patient-centred.
The principle we adopted was that expertise, information and standards should travel but that the patient should not have to. That shaped the development of Albania’s National Cancer Institute and National Cancer Network. We linked university, regional and local services supported by a strategic partnership with the Istituto Nazionale dei Tumori in Milan.
The NHS begins from a vastly stronger base, but scale creates its own difficulties. World-class expertise is of limited use if it remains trapped behind institutional walls. A centre of excellence is important, but a system of excellence, while harder to build, is ultimately far more valuable.
The hardest part of healthcare innovation is not inventing the technology, it is making it work for patients (Getty/iStock)
My experience as a co-founder of Lucida Medical, a UK company emerging from academic work in AI and cancer imaging, taught me something similar. Developing an idea is difficult. Getting it into clinical practice is often harder. Technology arrives in a world of competing priorities and, inconveniently, human beings. Government simply made that reality impossible to ignore.
This is why I have become wary of announcing success too early. Success is not an algorithm deployed. It is an aggressive cancer treated at the right moment. It is an unnecessary biopsy avoided. It is specialist expertise reaching someone who does not happen to live beside a famous hospital. It is a doctor having the right information at the point of decision; a patient who feels cared for rather than processed.
For all my enthusiasm about AI, medicine must remain human. AI can bring together imaging, clinical history and medical evidence at a speed that no individual physician could hope to match. But an orchestra still needs a conductor. The doctor must remain responsible for context, uncertainty and the individual sitting in front of them. This matters especially in cancer, where decisions may involve a multitude of factors none of which fits neatly into an algorithm.
If AI can remove repetitive work and assemble information intelligently, clinicians may have more time for judgement, explanation and human attention. The future was never really about human vs AI, but how to make better humans by working with better technology.
For most of my professional life I have inhabited the rather more controlled world of academic medicine: cancer imaging, research, clinical trials and, increasingly, artificial intelligence. I have worked in Cambridge, New York and Rome, in institutions where the question is usually how to make medicine better. I now return to Gemelli hospital in Rome, from politics back to patients.