AI strategy in healthcare: the end-2026 deadline

Dutch healthcare is under pressure from two directions at once this year. On one side, Zilveren Kruis announced that, as the first major health insurer, it expects care providers (GP care, district nursing, mental health care, hospital specialist care) to have a demonstrable AI strategy by the end of 2026. In the Dutch system, insurers buy care from providers through annual contracts, so their expectations carry real weight. On the other side, enforcement of the EU AI Act (the European regulation on artificial intelligence) is approaching: from 2 August 2026 the Dutch Health and Youth Care Inspectorate (IGJ) will supervise the sector. If you still think AI in healthcare is an internal IT project that can wait until next year, you have not read those two signals carefully.

This article explains what healthcare organisations need to arrange in concrete terms, and why the urgency is greater than most boards realise. The AI Act part applies to care providers across the EU, including Belgium and Luxembourg.

Two forces at once: insurers and legislation

The Zilveren Kruis requirement is notable because it is not advice but a purchasing expectation. That means a provider without an AI strategy risks a less favourable contract position. And Zilveren Kruis is probably not the only insurer to take this line: market watchers expect similar signals from other large insurers in the coming quarters.

At the same time, the EU AI Act is coming into force in phases. For healthcare organisations two moments are crucial:

• 2 August 2026: Article 4 becomes fully enforceable. Organisations must be able to show that staff who work with AI systems are sufficiently AI literate. IGJ has been designated as the sector supervisor.
• December 2027 / August 2028: obligations for high-risk AI systems take effect (this has been pushed back somewhat through the Digital Omnibus). But the inventory of which systems are high-risk needs to happen now.

In healthcare, much more counts as “high-risk” than boards usually realise: diagnostic AI, treatment advice AI and systems for staff assessment in principle fall under Annex III of the regulation.

The infrastructure trap: 93% get stuck on legacy IT

There is a fundamental tension in healthcare that frustrates strategic ambitions again and again. The SOTI Healthcare Mobility Report (n=150 Dutch IT decision-makers in healthcare) shows that 93% of respondents say outdated systems hold back the introduction of new technology. Almost six in ten work with non-integrated systems that do not exchange data.

This explains an alarming figure from the Dutch Federation of Medical Specialists: more than 98% of AI tools introduced in healthcare never reach structural implementation. They end up in pilots, disappear after the test phase, and leave frustrated staff behind. Not because the technology is poor, but because the organisational and technical conditions are missing.

Outdated systems: 93% of Dutch healthcare organisations see legacy IT as a barrier to AI implementation. Photo: César Badilla Miranda / Unsplash

The “Realisation of AI in healthcare” programme of the Dutch Ministry of Health, Welfare and Sport (VWS) targets exactly this problem: €400 million in 2027 and another €400 million in 2028 from the Supplementary Care and Welfare Agreement (AZWA) are available for healthcare organisations that want to implement AI structurally. The first focus is on voice-driven reporting and capacity planning: two applications that tie directly into the government’s ambition to halve administrative time before 2030.

But that money is not for everyone. Healthcare organisations that want to claim it must show they have a strategy. Once again: the AI strategy is the key.

What already works, and why is the use of diagnostic AI falling?

Current AI adoption in Dutch healthcare shows a nuanced picture. According to the SOTI Healthcare Mobility Report, the most common uses are medical data processing (50% of healthcare organisations), administrative support (50%) and updating records (42%). These are relatively safe, low-hanging-fruit applications.

Striking is the fall in the use of diagnostic AI, from 48% in 2024 to 32% in 2025. The same goes for personalised treatment: from 52% to 31%. Not because the technology got worse, but because healthcare organisations pulled back after a hasty rollout without governance and enough training. The pattern is familiar to anyone who follows the literature: fast adoption without organisational embedding leads to disillusion and withdrawal.

That makes it all the more telling that a prototype at Veilig Thuis (the Dutch advice and reporting centre for domestic violence and child abuse, an organisation with limited resources and a high workload) achieved 50% time savings on administrative tasks with an AI assistant. The difference is not in the technology, but in the implementation approach.

What an AI strategy for healthcare must contain at minimum

An AI strategy is not a vision document full of fine words about innovation. For healthcare organisations that want to be taken seriously by insurers and regulators, it must contain at least the following elements:

An AI strategy starts with the professional: what does this mean for my daily work with patients? Photo: Nappy / Unsplash

1. An inventory of existing and planned AI applications Which tools are already in use, including by individual staff (shadow AI, meaning the use of AI tools outside official IT channels)? Which might fall under the high-risk classification of the EU AI Act? Without an inventory there is no strategy.

2. A plan for staff AI literacy Article 4 of the EU AI Act requires organisations to ensure sufficient AI literacy among staff who work with AI. That goes beyond a one-off workshop. It means measuring a baseline, differentiating training by role, keeping records and providing refresher training.

3. Governance and a responsibility structure Who is responsible for AI decisions? How are AI outcomes validated by a professional? How are incidents reported? IGJ expects healthcare organisations to have an answer to these questions.

4. An IT roadmap for integration As long as AI tools do not connect to the electronic patient record (EPD) or other core systems, pilots stay pilots. The strategy must address the infrastructure question, uncomfortable as that may be.

5. A link to quality of care and outcomes The most convincing AI strategy links technology to measurable care outcomes: less administrative time, shorter lead times, more reliable diagnostics. That also turns it into a concrete conversation with insurers.

"In healthcare organisations we often see that the technology has been there for a long time, but nobody has thought about what needs to happen next. Who trains the staff? Who monitors the quality of the AI output? Who is responsible if it goes wrong? An AI strategy answers those questions before you go live, not after."

Bram Kreuger, AI consultant & trainer, MSc AI, AI Advice & Training

The demographic argument: AI is no longer a choice

There is a long-term argument that puts today’s urgency in a different light. If demand for care in the Netherlands keeps growing over the coming decades (and it will), then by 2050 almost one in three working people would need to work in healthcare to keep the current ratio. That is demographically impossible.

So AI is not just a way to be more efficient. It is a structural precondition for a sustainable healthcare system. Healthcare organisations that invest now in a sound AI strategy are building their own resilience for the future, not just compliance.

What to do now

The end-2026 deadline sounds comfortable, but it is not. An AI strategy that convinces both Zilveren Kruis and IGJ takes time: taking inventory, building support, setting up governance, drawing up training plans. Organisations that only start in the autumn of 2026 will not make it.

Our recommendation: start the inventory phase in Q3 2026, appoint an internal lead (Chief Information Security Officer (CISO), Chief Medical Information Officer (CMIO) or a dedicated AI coordinator), and make sure the executive board recognises this as a strategic priority, not an IT project.

We work with healthcare organisations on exactly this journey: from baseline measurement to governance design and AI literacy training for staff at every level. Get in touch if you want to know where your organisation stands today.