“Whether AI will change healthcare is no longer up for debate.”

Aleksandar Ivanovski is a senior machine learning engineer at Recare, a Berlin-based company that has been developing digital discharge management for hospitals since 2017 and now connects more than 700 acute care clinics. With his book “Beyond the Hype: A Practical Guide to AI Implementation for Business Leaders” (July 2025), he addresses decision-makers who not only want to understand AI, but also want to put it to practical use.

In an interview with #ai_berlin, Ivanovski talks about why Berlin is crucial to Recare's AI strategy, where the biggest gaps between AI promises and clinical reality lie, and how approaches such as federated learning and edge computing make data protection and innovation in healthcare compatible. A conversation about the courage to implement, shortly before the Recare AI Summit on April 20, 2026, in Berlin.

Mr. Ivanovski, Recare has been based in Berlin since 2017 and has become the leading provider of digital discharge management here. What role does Berlin play in AI development at Recare – and what makes the city special in your view?

Berlin has been a crucial location for Recare since its founding – both for our growth and for our AI strategy. The city is one of the most exciting tech and engineering centers in Europe and attracts outstanding talent, researchers, and entrepreneurs. This concentration of expertise and new ideas creates an environment in which innovations emerge quickly and are developed further at a high level.

Berlin is also particularly relevant for us because health intelligence is one of the clearly defined AI priorities here. The ecosystem is working intensively to apply AI in hospital processes, diagnostics, and patient care—precisely the field in which we are also active.

And finally, the location plays a very pragmatic role: Proximity to decision-makers, hospital networks, and public institutions has made it easier for us to build the trust necessary to responsibly anchor AI in one of the most heavily regulated industries.

Your book is titled “Beyond the Hype.” In your work at Recare, you experience firsthand the gap between AI expectations and reality in the German healthcare system. Where do you see the biggest discrepancy between what is promised about AI in the healthcare sector and what is actually feasible?

It was precisely this tension that prompted me to write Beyond the Hype. In healthcare, the challenge lies less in what AI can do in principle and more in the structures it encounters. AI needs interoperable, structured data – in reality, hospital IT is highly fragmented, interoperability is limited, and much of the information is unstructured. Many organizations fail not because AI offers no added value, but because their infrastructure cannot support it.

Our approach at Recare is therefore different. We integrate our technology directly into existing systems and processes – both structured and unstructured data. This makes AI usable in everyday life without clinics having to rebuild their IT infrastructure. The real challenge is not intelligence, but implementation. Closing this gap is our focus.

In your book, you describe federated learning and edge computing as promising approaches to data protection in healthcare. How does Recare implement these principles in concrete terms – and what technical challenges did you have to overcome in the process?

Federated learning requires a fundamental rethink: instead of bringing data to the model, you bring the model to the data. That sounds theoretical, but it can be implemented in a very concrete way – and that's exactly what we've done at Recare.

With Recare Predict, we don't ask hospitals to send us sensitive patient data. Instead, we bring the prediction model directly to the clinic, where it runs entirely within the existing infrastructure. This is where edge computing comes in: the model learns from local processes and patient flows and optimizes discharge management – without a single data point leaving the premises.

This is technically challenging. Different IT landscapes and limited computing capacities make rollout and maintenance complex. To address this, we have established stable validation processes and close feedback mechanisms that enable continuous development of the model without centralizing sensitive information. Recare Predict shows that privacy by design is not an obstacle, but a quality feature. Consistently implemented, it strengthens trust and facilitates adoption.

Recare now connects 700 acute care hospitals and over 24,000 follow-up care providers. In your book, you warn that the biggest challenges in scaling AI are often organizational in nature. What were the specific stumbling blocks in Recare's growth – and how did you overcome them?

Scaling AI in healthcare means tackling two very different challenges in parallel. Internally, the main issue was to exercise restraint. There is a great temptation to use AI in as many areas as possible at the same time. We therefore created structures in which all teams regularly reflect on what creates real added value and what is not yet viable – not only in engineering, but throughout the entire company. This allows each department to decide for itself where AI makes sense and where it does not.

Externally, the starting point is different. Hospitals are complex organizations with established system landscapes, scarce resources, and high regulatory pressure. That's why we decided early on to develop our solutions in such a way that they fit into existing processes. Our technology adapts to the processes – not the other way around.

On April 20, 2026, the Recare AI Summit will take place in Berlin – with up to 1,200 participants under the motto “Rebooting the Health System.” What exactly can visitors expect, and what message would you like to convey to decision-makers in the healthcare sector?

The Recare AI Summit is not about buzzwords, but about concrete implementation. The stage will feature real projects, tangible results, and best practices from organizations that are already successfully using AI. Participants can expect a practical program, honest exchange, and, above all, clarity about what it really takes to integrate AI securely, scalably, and sustainably into clinical processes.

My message to decision-makers is clear: whether AI will change healthcare is no longer up for debate. The decisive factor is how it is introduced strategically and sensibly. This is exactly what the Recare AI Summit provides guidance on.

A central message of your book is: “ROI comes from the value that AI enables, not from the technology itself.” Can you give a concrete example from your work at Recare where AI has created measurable business value?

This attitude shapes our product development at Recare. Two examples illustrate this concretely: Recare Predict focuses on discharge management. Social services often work reactively: only when treatment is complete does the search for a suitable facility begin – which leads to unnecessarily long hospital stays. Recare Predict recognizes during the course of treatment whether a transfer is likely and makes it possible to initiate the next steps in good time. The result: measurably shorter stays, less stress for patients, and noticeable relief in terms of beds and capacity.

Recare Docs addresses the administrative burden: doctors invest around three to four hours a day in documentation, discharge letters, and organizational handover. Recare Docs automates these processes, saving numerous hours per week and per clinic, making discharge letters available more quickly, and ensuring that patients receive their documents without unnecessary delays. In a system where time is one of the scarcest resources, this creates measurable economic benefits and improves the quality of care.

The #ai_berlin hub brings together startups, research, business, and politics. In your opinion, which collaborations in Berlin's AI ecosystem would be particularly valuable for AI in healthcare – and where do you see Recare in this network in the coming years?

From a research perspective, a lot of relevant work is happening in Berlin – in application-oriented areas such as the optimization of hospital resources, medical language understanding, and AI-based decision support. Cooperation is crucial to bring these developments into clinical reality – where they can be tested in everyday life. This is precisely where Recare can play an important role.

Regulatory frameworks must take into account the actual processes in everyday clinical practice – which can only be achieved through close cooperation between policymakers and the companies that use such systems operationally. And interoperability remains central: Companies working on related issues will gain more in the long term from common standards than from isolated individual solutions.

I see Recare playing two roles here: as a connecting element that brings together innovation and healthcare practice with more than 1,000 acute care hospitals and 26,000 nursing and rehabilitation facilities – and as a voice for implementation-oriented AI. We talk openly about what works, where the limits lie, and what it takes for the industry to move forward. We embrace this responsibility – in Berlin and beyond.

Thank you very much for talking to us.

Note: This interview was originally conducted in German and subsequentially translated into English language.