Healthcare AI you can trust: What it really takes

Insights from Nordic healthcare leaders on evidence, transparency and clinical judgment

When a clinician turns to AI for support at the point of care, one question matters above all else: can they trust the answer and that it's grounded in verifiable evidence, traceable to its source and solid enough to support a real decision about a real patient? That's the bar clinical AI has to clear — and it's the question that shaped a candid dialogue between Nordic clinical leaders in Oulu. 

Co-hosted by Tim Morris, VP Commercial at Elsevier, and Pooja Jha, Editor-in-Chief of The Lancet Regional Health - Europe, the gathering brought together CMOs, professors of neurology, health informatics specialists and district chief physicians. The conversation wasn't a debate about whether AI belongs in clinical settings, but rather an examination of what it takes for AI to earn a place there. Four priorities emerged: trusted content as the foundation, transparency in how AI reaches its conclusions, the non-negotiable role of clinician judgment and a shared responsibility for responsible adoption. For Elsevier, those priorities aren't abstract. They reflect the basis on which meaningful partnerships with health systems and academic institutions are built. 

Trusted content as the foundation 

If confidence in AI has a foundation, it's evidence. Clinicians don't simply want an answer, they want to know where it comes from. An AI response that can't point to its source asks for a leap of faith that few healthcare professionals are willing to take. 

When a clinician acts on an AI response, the evidence behind that answer must be verifiable, current and drawn from credible sources. The Nordic region's strong governance culture and high levels of public trust in health institutions reflect a broader standard — one that applies equally to the AI tools clinicians use at the point of care. Reliable data and reliable evidence are the same principle, applied at every level of the system. 

That's why approaches that link AI outputs to traceable, retrievable sources drew consistent support in Oulu. Not as a technical preference, but as a matter of accountability. ClinicalKey AI is built on this premise: every answer is traceable to Elsevier’s peer-reviewed clinical library. The value is not just speed — it is confidence. 

Transparency you can follow 

Closely tied to evidence is transparency. Clinicians want to understand how a tool reaches its conclusions and what it draws from. That's not a technical preference — it's a professional standard, and it reflects how good medicine already works. A diagnosis is supported by reasoning, a treatment plan by guidelines and a referral by a clear rationale. The same expectation applies to AI. Tools like ClinicalKey AI, which link answers directly to supporting published evidence, offer one practical example of what that transparency looks like day to day. 

Clinician Judgment, the Non-Negotiable 

One principle held firm throughout the Oulu discussion: the clinician decides. AI can inform, surface evidence and support reasoning, but the final call belongs to the professional in the room. That boundary wasn't presented as a limitation — it was presented as a condition for confidence in the tool. 

The group was unanimous that human oversight isn't a preference; it's a requirement. And with that came a pressing question: are institutions building the skills to make oversight meaningful? AI literacy — the capacity to critically evaluate outputs for flaws, bias and hallucinations — was identified as one of the most urgent institutional priorities. The question the group kept returning to was whether critical appraisal, a cornerstone of clinical education, is being taught with AI in mind. 

The consensus wasn't about constraining AI. It was about keeping the human in the loop: not a clinician who defers to the machine, but one who is better equipped to think, question and decide. 

The Nordic advantage 

The Nordic region is well positioned to help shape what responsible clinical AI looks like in practice — not because of any single advantage, but because of values already embedded in its health systems: rigorous governance, strong institutions and a culture of collaboration that extends across borders.1 

These qualities matter because good clinical AI doesn't start with technology. The Nordic region's existing commitment to those values is a meaningful enabler of what responsible AI can look like at the point of care. 

Realizing that potential involves ongoing investment: better integration across data sources, sustained commitment to infrastructure and talent, stronger public-private partnerships and deeper cross-border collaboration. As the dialogue in Oulu made clear, these aren't future aspirations. They're the practical conditions that allow evidence-backed AI to function at the level clinical care demands. 

A path worth building together 

What emerged from Oulu was not a set of objections to AI, but a set of standards for it. Lead with verifiable evidence. Keep reasoning visible. Hold clinician judgment at the center. Those aren't aspirational principles — they're the conditions clinical AI must meet to earn a place in care. These are also the conditions that Elsevier is committed to supporting, through evidence-linked clinical tools, rigorous evaluation frameworks and a publishing infrastructure built on scientific integrity. 

The same commitment applies in research. AI adoption is no longer a future scenario — it's current practice. Currently, 55% more biology papers are now written with AI; 84% of researchers actively use it; and 53% of peer reviewers do too. 2  

The right response isn't alarm, but accountability. Scientific standards haven't changed, and as AI becomes more routine in research and publishing, expectations around disclosure and integrity only become more important. Emerging publishing norms, including AI disclosure requirements and the editorial infrastructure that supports them, are a meaningful part of that framework. 

None of this happens in isolation. It takes clinicians, researchers, publishers and health systems willing to hold each other to the same rigorous expectations. The Nordic region offers a strong example of what that kind of shared commitment can look like. But the work reaches well beyond any single region. It takes partners who bring both depth of evidence and a genuine infrastructure of accountability. That's what Elsevier is here to support: not just tools, but a foundation of knowledge that clinicians can rely on at every point where a decision matters.