AI in healthcare: not a replacement but support

"AI is going to revolutionise healthcare" has been the conference line since 2023. The reality in a GP practice in Zeeland is calmer: a tool that turns a consultation note into a structured report saves 90 minutes a day. Not glamorous, but it works. The spectacular diagnostic AI from the press release is often still in pilot, while the boring administrative AI already hands patient time back.

That is the honest picture. Not utopia, not collapse. What does it mean for patients, doctors, and Dutch healthcare?

What does AI mean for healthcare professionals?

AI mostly changes the boring part of the work. Note-taking, coding, triage, callback protocols. That is where doctors and nurses lose time, and where AI delivers value without much clinical risk. Diagnostic support is possible, but a different conversation with very different validation requirements.

The challenge is not the tech, the tech works. The challenge is integration: linking with the electronic patient record (HiX, Epic, Nedap), GDPR-compliant data handling, and clinicians actually using the tool. Not a technology problem, an implementation problem.

Transparency

One thing is non-negotiable: you must be able to see how an AI reached its conclusion. A black box that says "elevated risk" without explanation is not a tool in healthcare, it is a liability. For administrative AI the bar is lower, for diagnostic AI explainability is a hard requirement.

Professor Carl Moons of UMC Utrecht says: "Artificial intelligence has enormous potential to improve patient care and preventive care, across all healthcare sectors. From early diagnosis of patients with lung cancer to identifying those at increased risk for heart attacks, dementia, birth defects, infections and many other conditions."

AI is not a replacement, but it is a tool

AI does not replace doctors. Not because the tech could not, but because healthcare needs a person who carries responsibility, reads context, and can stop when something feels off. That is part of the job.

And yes, AI makes mistakes. Sometimes different mistakes than a human would, which is what makes it tricky. That is why a clinician stays in charge and can correct those errors.

Eric Topol, cardiologist and AI expert, makes this point in his book "Deep Medicine". He argues that AI can not only contribute to better diagnoses, but also improve the human side of care by automating admin work. The second effect is the bigger short-term win.

How is AI already being used in healthcare?

Diagnostics and treatment

A few examples are already running in the Netherlands. The WoundHealth app from Radboudumc lets patients photograph their wounds at home; AI checks whether the wound is healing properly. In radiology, AI systems help analyse X-rays and CT scans, allowing radiologists faster and more accurate diagnoses. None of these tools work without a clinician behind them, and that is how it should be.

A study published in Nature (2020) showed that an AI system by Google Health analyzed mammograms to detect breast cancer, outperforming human radiologists in some cases.

Preventive care with AI

AI systems can analyse large amounts of health data to identify risk factors and enable early interventions. The Mount Sinai Health System in New York developed the "Deep Patient" AI platform that analyzes electronic medical records to make predictions about disease outcomes. Useful only if you know how to act on those predictions inside a clinical workflow.

Predictive care in practice

A concrete example is Nicolab's StrokeViewer. This AI-driven platform enables physicians to diagnose strokes faster and more accurately by applying advanced image analysis algorithms to CT scans. Time saved here is direct clinical benefit, because in a stroke every minute counts.

Challenges of AI in healthcare

In the Netherlands, healthcare AI lives or dies on three things: data quality and privacy (GDPR plus Wbga for genetic data, plus NEN 7510), explainability, and liability when something goes wrong. The last one is not legally settled, and that is what holds many boards back.

On top of that, the practical reality: integration with the EPD. A tool that does not write back to HiX or Epic is, for most institutions, another dashboard nobody opens. Ask every vendor: does it write back, in what format, and who maintains that integration over time?

Limitations of AI

AI in healthcare is impressive, but not infallible. A pattern recogniser does not grasp the full context of a patient: home situation, prior complaints, what the partner said on the phone. AI shows no empathy and makes no ethical decisions. Not a temporary gap that more data closes, a fundamental difference. A clinician remains indispensable.

The future of AI in our healthcare system

The future is not "AI does the diagnosis and the doctor clicks OK". A more realistic picture: AI does the prep work, the clinician decides, AI does the write-up.

Personalisation of treatments

AI can help tailor treatments to individual patients based on earlier treatments, lab results, and literature. This can lead to more effective treatments, fewer side effects, and higher chance of recovery. Not equally relevant for every condition, but concrete in oncology and chronic care.

More efficient care processes

By automating tasks, AI relieves clinicians of admin work, leaving more time for what matters: personal contact with patients. The biggest short-term win, structurally underrated because it is not spectacular.

Early detection of disease

AI systems can help detect disease early, sometimes before obvious symptoms. The practical roll-out depends on whether the system can absorb those signals, because early detection without follow-up capacity is not progress.

Practical AI in Dutch healthcare: where you can start today

Beyond the spectacular applications (radiology, protein structure, image analysis) there are equally valuable practical ones where Dutch healthcare institutions can make a difference today.

The biggest direct win is administrative relief: generating reports from consultation notes (Augmedix, Suki, Nuance Dragon) saves on average 1 to 2 hours per doctor per day, which flows directly into patient contact. Triage and scheduling is the strong second: AI systems that classify symptom input and refer to the right department, lowering wait times and wrong referrals at GP posts and emergency departments.

Patient communication runs in parallel: automated answers to standard questions (medication, appointments, preparation for examinations) via secure WhatsApp Business or patient portals, 24/7 without burdening staff. And finally documentation and coding: automatic DRG and ICD-10 coding and quality registrations from file content, saving admin time and improving data completeness.

GDPR, healthcare standards, and compliance

Healthcare data is particularly sensitive and falls under special GDPR categories plus the NEN 7510 standard for information security in Dutch healthcare. Cloud AI with patient data is only allowed under strict conditions: data processing agreement, EU-only storage, no training on client data, transparency to the patient. A vendor that cannot answer those four clearly is not a vendor for healthcare.

For the most sensitive client data (youth care, mental health files), on-premise or an EU-private cloud is a serious option. Discussed per case; not the default. That kind of setup can also help with healthcare inspection requirements.

Conclusion

AI in healthcare is not a revolution, it is a tool. The biggest gain is not in spectacular diagnostic AI, but in the boring layer underneath: admin, coding, triage, and patient communication. That is where time is already being given back to the people who actually deliver care.

Start small on a process that visibly hurts (often note-taking), build the EPD integration from day one, and make GDPR and NEN 7510 hard criteria in vendor selection. No two-year pilot, one tool live in one department, then expand.

For healthcare organisations that want to start concretely: a short analysis maps out which processes deliver the most value, with attention to GDPR, healthcare standards, and the culture of your institution.

Curious how AI can support your healthcare organisation? Take the free AI Scan for a personalised analysis within 1 minute.