Across Europe, hospitals are increasingly combining data platforms and artificial intelligence to improve patient care. At the same time, they must navigate growing regulatory requirements, privacy concerns and the debate over digital sovereignty. Belgian regional hospital AZ Delta is one of the organizations balancing these competing priorities. CIO Peter de Jaeger explains how the hospital is building a long-term data and AI strategy while preparing for an increasingly complex European digital landscape.
Earlier this year, AZ Delta announced a major investment in a new generation of AI-enabled healthcare technology designed to support clinical decision-making. At the heart of that strategy is a secure data platform capable of bringing together, enriching and analysing vast amounts of clinical information. The objective is to identify risks earlier, predict disease progression more accurately and help clinicians tailor treatments to individual patients.
Behind that announcement lies a multi-year transformation programme in which cloud technology, data governance, AI and collaboration are closely intertwined.
From Industry to Healthcare
Before joining AZ Delta in 2020, Peter de Jaeger built his career in the steel industry, where data already played a central role in business operations. Healthcare, however, proved fundamentally different.
"The biggest change was realizing that data is no longer simply supporting the organization," he says. "Today, it has become one of the foundations on which healthcare operates."
The COVID-19 pandemic accelerated digital transformation across hospitals. Now, according to De Jaeger, the challenge is no longer digitisation itself, but embedding technology into everyday clinical practice while addressing new questions around generative AI, cybersecurity, privacy, resilience and digital sovereignty.
Building a Central Data Platform
Rapid digitalisation and explosive data growth convinced AZ Delta that its fragmented data landscape had become unsustainable.
"When I arrived, information was scattered across different environments," De Jaeger explains. "Some systems were on-premises, others were already running in the cloud. Different departments maintained their own datasets, researchers had their own copies, and individual physicians often managed data independently. Whenever data changed, it frequently had to be updated in several different places."
The hospital therefore decided to consolidate virtually all clinical data into a single cloud-based platform built on Google Cloud. Every 24 hours, data from the hospital's operational systems is securely transferred through an on-premises gateway into the cloud environment.
To organise that information, AZ Delta uses a medallion architecture, separating datasets into bronze, silver and gold layers according to their quality, sensitivity and intended use. "Privacy and security are fundamental," De Jaeger says. "The advantage of this architecture is that data only has to be prepared once—for example through pseudonymisation—and can then safely be reused for multiple purposes."
Managing 280 Billion Data Points
Building the platform required migrating approximately 280 billion data points, with millions of new records being added every day. The challenge was not only storing that information but making it immediately accessible.
Clinicians require rapid access to patient records during consultations, while researchers and AI developers need to analyse enormous datasets for model development and scientific studies.
"If you submit a query, you don't want to wait two hours for an answer," De Jaeger says. "Especially not when you're in the middle of a clinical or analytical thought process." Performance therefore became a key design principle.
Google Cloud offered the hospital sufficient computing power while allowing a strict separation between production environments, such as the electronic health record (EHR), and secondary environments used for research, analytics and AI development. That separation ensures research workloads never interfere with day-to-day clinical operations.
Balancing Performance and Digital Sovereignty
The decision to build the platform on Google Cloud inevitably raises questions about Europe's ambition to reduce dependence on non-European technology providers.
De Jaeger acknowledges the debate. "No alternative platform currently offers the same capability to process large-scale datasets at this speed, while also providing such strong support for open-source software development," he says. "That combination is particularly important when you're developing AI."
He also recognises that geopolitical developments have made the discussion around sovereignty increasingly urgent. "We signed our agreement before digital sovereignty became such a dominant topic. Contracts like these aren't easily replaced overnight. But for the longer term, it is certainly something we continue to evaluate."
AI Beyond Radiology
With the underlying data infrastructure in place, AZ Delta has expanded its AI ambitions well beyond traditional imaging applications. Many commercially available AI solutions are already supporting radiologists, including software that automatically segments and grades prostate lesions.
At the same time, the hospital is increasingly developing its own AI models using internally generated clinical data. De Jaeger describes the approach as "engineer meets physician."
"We have the clinical expertise, we have the data, and we bring those together with software developers," he says. "Several applications are already moving through the development pipeline in areas including prostate cancer and cardiology."
One project has already attracted particular attention. Rather than relying on new diagnostic equipment, researchers are using two decades of routinely collected ECG data to identify patients at risk of cardiac amyloidosis, a rare disease that gradually stiffens the heart muscle and often remains undiagnosed until its later stages.
According to De Jaeger, approximately one in ten people over the age of 65 may develop the condition, yet only a small fraction are currently diagnosed. . Although treatment is available, diagnosis often comes too late.
"Our AI models can identify patients years before they would normally receive a diagnosis," he says. "In 2024 alone, we identified around one thousand people in our region who are likely to have the disease. On November 18, 2025, we were able to provide preventive treatment to the first patient identified through this approach. Today, every ECG performed in our hospital is analysed automatically."
For De Jaeger, the project demonstrates that AI's greatest value often lies not in replacing clinicians, but in uncovering clinically relevant information hidden within data hospitals already possess.
Avoiding Pilot Fatigue
While enthusiasm for AI continues to grow, De Jaeger warns against launching dozens of disconnected pilot projects. "Innovation depends on three things: people, data and funding," he says. "Those resources should be invested carefully. I'd rather focus on a limited number of projects and do them well than spread ourselves too thin."
That philosophy is captured in one sentence that has become something of a guiding principle within AZ Delta. "Innovation is a marathon, not a sprint."
Maintaining focus, he believes, creates the conditions needed to scale successful projects into routine clinical practice instead of allowing promising ideas to remain isolated experiments.
He also expects AI development itself to accelerate rapidly over the next few years. "Many university hospitals and regional hospitals are now building similar data platforms," he says. "Within two years, I believe this will become standard practice. Developments such as vibe coding and AI-assisted software development using tools like Claude are making digital innovation faster, more affordable and much more accessible."
Collaboration at National Scale
Although AZ Delta is considered one of Belgium's frontrunners in clinical AI, De Jaeger stresses that meaningful progress depends on collaboration rather than competition.
In 2023, the hospital became one of the driving forces behind the Data Capabilities initiative, receiving €1 million in funding to support the development of shared data infrastructure.
That investment resulted in the creation of the Federated Health Innovation Network, bringing together hospitals that can jointly develop AI models while allowing patient data to remain within each participating institution.
"We used the funding to establish the network together with nine other hospitals," De Jaeger says. "Today, a newer version already connects fifteen hospitals."
Rather than centralising sensitive patient information, the federated approach enables algorithms to learn from distributed datasets while data remains under local governance.
The next ambition is even broader. Belgium currently has several regional initiatives, including one led by University Hospitals Leuven (UZ Leuven). De Jaeger believes these networks should eventually converge into a single nationwide ecosystem.
"Every healthcare organisation that generates clinical data should be able to participate," he says. "A national federated data infrastructure would allow us to innovate faster, conduct larger-scale research and develop AI applications based on data from the entire Belgian population."
AI as Part of a Broader Data Strategy
For AZ Delta, AI is not a standalone technology programme but the logical next step in a broader digital transformation. The hospital continues to invest in a secure data environment that combines structured clinical information - such as laboratory results and medication records - with unstructured data including medical images and clinical notes.
Equally important, says De Jaeger, is maintaining control over governance. By combining secure cloud infrastructure, robust data governance and internally developed AI models, the hospital aims to ensure that innovation remains aligned with clinical needs, privacy requirements and long-term strategic objectives.
For healthcare organisations across Europe facing similar challenges, AZ Delta's experience offers a valuable lesson. AI is not simply another technology investment. Its success depends on the quality of the underlying data, the strength of governance, close collaboration between clinicians, engineers and IT specialists - and the patience to build capabilities over time.
