What AI in Healthcare Really Needs
Co-authored by By Claudio Silvestrin
January 20th, 2026
Artificial intelligence in healthcare is often presented in sweeping terms: a revolution waiting to happen, a solution to inefficiencies, even a new diagnostic partner. But for those of us who have spent years inside both the science and the systems, the story is more nuanced. We are, in fact, living through what might be called the second hype cycle of AI in healthcare. The first came with the breakthrough of deep learning. The second, unfolding now, is powered by generative AI and agentic systems. Both have created enormous expectations. The harder question is: what will it take for AI to actually deliver?
As Claudio Silvestrin put it: “There was the first hype cycle when deep learning came along. The early algorithms were trained on vast amounts of image data. And so they were also good at detecting things on images — lesions, cancer, fractures, brain bleeds. It was a good fit for diagnostics.”
This was the starting point: narrow, image-based applications that enhanced doctors’ eyes rather than replaced them. But today, the scope is far broader. Generative AI and language models can interpret clinical notes, summarise patient histories, and generate new hypotheses for research. Suddenly, AI feels like an all-purpose tool that could permeate every part of healthcare. And yet, as Claudio also cautioned, “We’re definitely at the very beginning stages of really making proper use out of it.”
Healthcare, unlike many other industries, cannot simply ‘move fast and break things.’ Regulation, data privacy, and the complexity of clinical workflows mean adoption lags. Even when the technology exists, integrating it into practice is a different matter. This is not about algorithms alone; it is about reshaping the culture of institutions built over decades. Pharmaceutical giants, for example, have accumulated data at unprecedented scale, yet their legacy structures make deploying AI slow and fragmented. In contrast, smaller AI-first biotechs, particularly in the US and China, can operate with agility, embedding AI into their DNA from the start.
Claudio drew a striking analogy here: “If you want to be successful with AI, you need to go all in. I see the parallel to electric cars in the automotive sector. It’s a huge challenge for established companies like Volkswagen to make the switch. Whereas companies like Tesla in the auto world, are built for it from the ground up.”
The lesson is clear. What healthcare really needs is leadership, integration, and cultural readiness. Senior decision-makers must understand AI as a strategic capability. AI teams must become an integral part of the organisation with representation on the executive level, not just bolted on to existing structures. And above all, AI must be seen not as magic but as a tool that works best in partnership with end-users such as clinicians.
As Claudio concluded in our discussion: “I have seen AI save lives by detecting pathologies on medical images that otherwise would have remained undetected. I would always want a doctor together with AI to look at my case. And I think most people, if they had seen what I’ve seen, would feel the same.” The future is not hype, but partnership, and that is how AI will deliver on its promise.
