AI and Healthcare: What Is Actually Happening, and What Should You Make of It?

I have always been slightly sceptical of technology promises in healthcare. Not because I do not believe in the science, but because I have seen how rarely the science arrives at the bedside with any meaningful value.

So when I came across three separate sources this week, all pointing in the same direction, I thought it was worth sitting with them properly. The Nebius, AstraZeneca, and NVIDIA whitepaper from BioTechX Europe 2025. A BeInCrypto piece on what Pfizer, Anthropic, and longevity researchers are doing with AI right now. And a podcast featuring biomedical gerontologist Aubrey de Grey and immunology professor Derya Unutmaz on what AI-driven drug development could mean for human lifespan.

Together, they paint a picture that is genuinely interesting; and, I think, worth being honest about where the evidence is solid and where ambition might be running ahead of the proof.

What AI is doing right now in clinical settings

Let’s start with the fact that is hardest to argue with. According to the World Health Organisation, the global healthcare workforce shortage is projected to exceed 11 million practitioners by 2030. This reflects existing training pipelines, retirement rates, and the growing burden of noncommunicable diseases (NCDs). The WHO reported that NCDs claimed more than 43 million lives in 2021 alone, accounting for three-quarters of non-pandemic deaths globally.

You cannot fix a staffing crisis by training more staff more quickly - the pipeline is too slow. So the question becomes, what else do you do?

One AI solution described by the Nebius whitepaper roundtable participant enables nurses to monitor up to ten times more patients by removing manual file review. The tool integrates into Electronic Health Records and generates a faster, comprehensive patient overview. I think this is the most defensible and useful application of AI in healthcare right now. It does not replace clinical judgement, but removes the administrative layer between a nurse and the patient.

Other deployments from the whitepaper include:

• AI-driven medical scribes that automate consultation notes, freeing clinician time for direct care.
• An AI pipeline that tracks tumour volumes over time, enabling faster assessment of treatment effectiveness than traditional methods (which the whitepaper notes can take more than six months).
• Voice agents deployed to conduct at-home pre-anaesthesia evaluations, reducing surgical waiting times by streamlining a step that previously required a clinical visit.

The Nebius whitepaper is a roundtable summary rather than peer-reviewed evidence, so I would characterise it as credible expert opinion from practitioners who are already using these tools. The use cases described are grounded in real deployments, and are practical and specific.only requiring you to believe that removing low-complexity, repetitive tasks from overstretched clinicians is a good idea - I think that’s a belief most of us already hold.

What the biggest players are betting on

While the roundtable focused on clinical workflow, the BeInCrypto piece this week showed something different: the largest pharmaceutical companies and AI labs are placing much bigger bets.

Pfizer CEO Albert Bourla confirmed in a Bloomberg TV appearance that the company is actively reviewing a new molecule its scientists generated using AI. This is not a pilot or a proof of concept - it sits within Pfizer's stated corporate strategy. The company has invested up to $350 million in PostEra since 2020 for AI-designed small molecules, and in January 2026, Pfizer announced a collaboration with the Boltz biomolecular foundation model team to refine open-source models on Pfizer's own internal data. Bourla put the clinical rationale plainly in a Yahoo Finance interview: "Once we know the target where we need to hit, we need a medicine to do that. And AI can design medicines and molecules that can fit that target much faster and better than our own thing." Worth noting separately: Pfizer Ventures has also backed VitaDAO, a blockchain-based organisation that funds longevity research, which signals Pfizer's belief in Decentralised Science and biotech innovation that stretches well beyond conventional drug discovery.

Anthropic, meanwhile, launched Claude for Life Sciences in October 2025: a version of its model designed for biopharma professionals covering scientists, clinical trial coordinators, and regulatory managers. Sanofi, Novo Nordisk, and AbbVie are among those already using it. Anthropic has also recently moved to acquire Coefficient Bio, a stealth-mode biotech, in a deal reported at approximately $400 million.

These are structural commitments from organisations that do not make them lightly.

The bigger claim: AI and the future of disease and ageing

Derya Unutmaz, an immunology professor who appeared on the BeInCrypto podcast alongside Aubrey de Grey, was direct about where he believes AI is heading: "Very soon it's going to be malpractice not to use AI in medicine." He predicted most diseases could be addressed within ten to fifteen years. Aubrey de Grey, who has spent decades on the science of ageing, discussed his concept of "longevity escape velocity": the point at which life-extension science advances faster than a person ages. He puts the probability of reaching that point for humans by the late 2030s at roughly 50%.

I find these claims genuinely thought-provoking. I also think they deserve a grounding note, and to their credit, both speakers acknowledge significant bottlenecks. Creating biological digital twins to model drug responses would, per the podcast, require roughly a thousandfold increase in current computing power, alongside vast datasets on the dynamic physical behaviour of cells at a level medicine does not yet have. Aubrey de Grey himself points to regulatory frameworks in Western healthcare systems as a possible brake on progress, even if the science delivers. The near-term milestone he treats as meaningful is not human longevity extension, it is robust mouse rejuvenation, which he sees as the proof of concept that would shift both societal and regulatory attitudes.

The ambition is real..the timeline is speculative. Both things are true, and I think being clear about that distinction matters, particularly for anyone making commissioning or procurement decisions based on what you read.

The barriers are not just technical

The Nebius whitepaper is candid about what is actually slowing AI adoption in healthcare, and this section resonates with everything I have seen in my own consulting work.

Three systemic problems keep emerging:

**Reimbursement models are outdated. **If an AI solution allows a nurse to monitor ten times more patients, but payment structures still reimburse per task performed by a human professional, the productivity gain is financially invisible. The roundtable participants were direct: modernising payment frameworks is essential, and insurers need to update reimbursement guidelines to reward outcomes rather than activity volume.

Procurement cycles are too slow. The whitepaper cites contracting timeframes of eight to twelve months for hospital AI procurement. In that window, the technology may have evolved, the startup may have run out of runway, and the clinical team who championed the project may have moved on. Pfizer's model of long-term committed investment (the PostEra relationship has been running since 2020) is a useful contrast to how most healthcare systems currently operate.

**Cultural adoption is underestimated. **Many clinicians fear being replaced. The roundtable's participants are direct about this: the solution is not to dismiss the fear, but to design for it. Role-specific training, early involvement of clinical opinion leaders in development, and honest communication about what AI tools can and cannot do all matter.

What I think

AI cannot, I think, replace human care. I believe that health and wellness is intrinsically linked to having human relationships; from being genuinely heard, from someone holding space with you when you are worried. That is not something a model should do, or something that can be substituted by a robot. What AI is doing, and what I believe it will do more of, is removing the structural weight that stops clinicians from being present with patients: the documentation, the manual reviews, the administrative overhead.

The longevity science conversation is fascinating, and I do not want to dismiss it. Pfizer's investment in PostEra, Anthropic's commitment to life sciences, and serious researchers making probabilistic predictions about disease timelines all suggest something real is forming. But I think the more urgent and tractable question right now is: how do we stop burning out the workforce we already have?

There is also a warning that I think gets insufficient attention: AI-driven complacency. Journalist Shane Harris recently asked Claude how it felt about being used in a US military targeting system. Claude's response included this: “...when a system generates hundreds of targeting recommendations and humans spend roughly the equivalent of a glance approving each one, the human is not really making a decision in any meaningful sense. They are ratifying an algorithmic output under time pressure with incomplete information and institutional pressure to move fast.” Claude called it "not human judgement..that’s automation bias with a human signature attached."

Overstretched clinical teams working under time pressure are exactly the conditions in which a clinician approves an AI recommendation without the scrutiny it deserves. Augmentation only works if the human remains genuinely and meaningfully in the loop (a requirement of the EU AI Act); the risk is that an AI system that appears effective on paper is precisely the kind that breeds complacency in practice.

The answer, based on everything I have read across these sources, is the combination of practical near-term AI dep loyment in clinical workflows, modernised payment structures, procurement reform, and the kind of multi-stakeholder collaboration that the Nebius roundtable called for. The Pfizer and Anthropic commitments show what sustained, strategically aligned investment looks like. Most healthcare systems are nowhere near that level of clarity yet.

The question for commissioners, policymakers, and healthcare executives is not whether to engage with AI - that decision has already been made, whether you were at the table or not. The question is whether you shape it with enough clinical grounding to make it genuinely useful, and whether your procurement and incentive structures are capable of keeping up.