.webp)
AI dominated the conversation at HLTH Europe this year. But beneath the excitement around new tools, models and announcements, a more nuanced discussion was taking place.
The most interesting conversations weren't about whether AI has a role in healthcare – that question feels increasingly settled. Instead, speakers repeatedly returned to a different set of challenges: who is accountable? How do we implement it? And where do clinicians sit in this AI-enabled future?
Here are six themes that stood out.
1. Accountability will define the next phase of AI adoption
If there was one topic that consistently resurfaced across sessions, it was accountability.
Healthcare already has established frameworks for determining responsibility when traditional medical devices fail. If a defective imaging machine contributes to patient harm, there are clear processes for investigating whether responsibility lies with the manufacturer, healthcare provider or user.
AI introduces a new layer of complexity. If an AI tool generates an incorrect recommendation that influences a clinical decision, where does accountability sit? The clinician? The healthcare organisation? The AI provider?
One audience member at the ‘L-AI-bility – who is responsible when it goes wrong?’ panel discussion captured the challenge succinctly, arguing that responsibility for AI errors should sit with the provider rather than the clinician, just as manufacturers are held accountable when traditional medical devices fail.
Until the industry provides clearer answers around liability, governance and responsibility, trust will remain one of the biggest barriers to widespread adoption.
2. Healthcare doesn't have an AI problem, it has an implementation problem
One of the clearest messages from HLTH was that the tech is no longer the biggest barrier.
Clinicians are adopting AI scribes at pace. Patients are increasingly turning to AI for health information. New solutions are entering the market every week. Yet many healthcare organisations remain stuck between promising pilot programmes and meaningful transformation.
Several speakers pointed to the same issue: implementation. Success depends on far more than choosing the right tool. It requires clinician buy-in, governance, AI literacy, workflow redesign and clear ownership. Dr Hannah Allen, CMO at Heidi, said their approach to deployment, for example, starts with compliance and education before rolling out department by department with clinical champions.
As organisations rush to explore AI, the conversation is shifting from capability to change management. As one speaker argued, healthcare may need fewer data scientists and more implementation specialists. The challenge is no longer proving that AI can work. It's embedding it into everyday clinical practice in a way that is safe, scalable and sustainable.
3. Clinical judgement remains healthcare's most valuable asset
Throughout discussions on AI-enabled care, speakers repeatedly emphasised that technology should support clinical decision-making, not replace it.
During a session on AI and clinical training, Erika Doyle-Hall, clinical advisor to Heidi, highlighted the opportunity for AI to reduce cognitive burden and remove many of the repetitive tasks that contribute to clinician burnout. But she was equally clear about where human expertise remains essential.
"Some things will never change," she said. "The communication between patient and clinician [is paramount]." As AI becomes more embedded in healthcare, the ability to apply judgement, navigate uncertainty and communicate effectively with patients may become even more valuable.
That perspective surfaced repeatedly throughout the event. Speakers pointed out that patients don't simply need information; they need context and guidance. During the Oxford-style debate, cardiologist Michael Winter argued that clinicians help patients navigate uncertainty and make difficult decisions in ways technology cannot. Whether it's discussing treatment options understanding an individual's circumstances, healthcare is rarely a matter of applying information alone.
Perhaps the biggest challenge for healthcare organisations isn't teaching clinicians how to use AI, but ensuring they don't stop exercising their own judgement. One speaker during the ‘Hallucinations or Misplaced Creativity: Is AI Improving Clinical Training?’ talk referenced a Mayo Clinic study showing clinicians occasionally overruled their own instincts and accepted incorrect AI recommendations. As AI becomes more embedded in healthcare, the most effective clinicians may not be those who rely on it most heavily, but those who know when to trust it and when not to.
4. Relationships are not an inefficiency
When this Oxford-style debate asked whether big tech will ultimately deliver better healthcare than hospitals, advocates argued that technology can improve access, scale prevention and create more personalised experiences. But the discussion repeatedly returned to a central question: can healthcare be reduced to data and algorithms alone?
Speaking against the motion, Nikita Kanani, of preventative healthcare screening service (and Semble customer) Neko Health, argued that the value of healthcare lies not just in the information, but in interpretation, trust and continuity. As she put it: "I spend time with patients and build a relationship over time. Tech cannot replace what I offer."
While AI and digital tools can support decision-making, patients still need clinicians to help them navigate complexity and uncertainty. "People use the chatbots and then come to see their physician for the translation," Kanani noted, highlighting the continued importance of human expertise in making sense of increasingly sophisticated health data. (You can read Semble's own research on AI-powered patients right here.)
For healthcare leaders, the debate reinforced a theme that surfaced throughout HLTH: the future is unlikely to be hospitals versus technology, but hospitals and technology. Digital tools may transform how care is accessed, coordinated and delivered, but clinical relationships remain fundamental.
In Kanani's words, "Relationships are not the inefficiency, they are the whole point of healthcare – let's not price them out of the equation." As providers invest in AI and data-driven care models, the challenge will be using technology to strengthen human-centred care rather than replace it.
5. New models of care are becoming a necessity, not a choice
Few topics generated as much discussion at HLTH as GLP-1s and the growing demand for obesity treatment.
Speaking during a panel on the rise of weight-loss medications, Earim Chaudry of Voy challenged some of the assumptions that still surround obesity care. Many patients arriving at Voy have already spent years trying to manage their weight through traditional approaches, often with multiple failed attempts behind them. Yet stigma remains a major barrier, with more than half of patients reportedly hiding their treatment from their partners.
The discussion quickly moved beyond the medication itself. If obesity treatment is becoming more accessible and demand continues to grow, can traditional healthcare models keep up? Chaudry argued that systems built around face-to-face consultations were never designed to serve this volume of patients. "We need to really look deeply at new models of care in order to serve people," he said.
That may be the bigger lesson from the GLP-1 boom. Beyond the clinical outcomes, these treatments are exposing a broader challenge for healthcare: how to deliver preventative, long-term care at scale. As more people seek support earlier in their health journey, providers will need new approaches that balance accessibility, continuity and clinical oversight.
6. Europe's AI advantage may be trust, not speed
Much of the global AI conversation is dominated by comparisons with the United States. But several speakers argued that Europe has strengths of its own.
Lukas Saari of Tandem highlighted in ‘Europe’s play in the AI race’ that Europe's focus is on safety, regulation and localisation. Building healthcare technology across multiple languages, healthcare systems and regulatory environments is undoubtedly challenging, but it also creates products that are often better suited to real-world clinical settings.
As Lukas noted, Europe is often better at ensuring products are safe, while the US tends to move faster when adopting new technology.
In healthcare, where trust is essential and mistakes carry significant consequences, that focus on safety and clinical validation may prove to be a competitive advantage rather than a limitation.
What's next for healthcare?
The biggest takeaway from HLTH Europe wasn't that AI is coming to healthcare. It's already here.
The more important question is what happens next.
Healthcare leaders are now moving beyond conversations about capability and towards questions of implementation, accountability and trust. The organisations that succeed will be the ones that deploy it responsibly, integrate it into clinical workflows and use it to strengthen, rather than replace, the human side of care.
