Among the world’s top 10 smart hospitals, only one is based in Europe: Charité – Universitätsmedizin Berlin, ranked eighth by Newsweek and Statista. Dr. Peter Gocke, Head of Digital Transformation at Charité – Universitätsmedizin Berlin, discusses with ICT&health in an interview in two parts what defines a smart hospital beyond just going paperless, shares how Charité is navigating AI adoption with over 70 ongoing projects, and explains why structured data is more valuable than cutting-edge hardware.
As a patient in a hospital, how would I notice that I’m not in a regular hospital but in a smart hospital?
“You’d notice it by how well the processes function. There would be fewer waiting times, and you wouldn’t have to provide the same information repeatedly. In hospitals where information flows digitally and comprehensively, the whole environment feels different, almost as if there are fewer patients.”
“That’s because many inefficient steps, like sending people to fill out forms, chase down paperwork, or retrieve patient records manually, are eliminated. A key indicator is that the information you once submitted is readily available everywhere, eliminating the need for staff to repeat or physically fetch it.”
So the less visible paper, the smarter the hospital?
“Exactly, less visible paper and fewer people having to run around gathering information. When information retrieval involves physical movement, it's a sign the system isn’t yet smart.”
And how does the staff recognize that they’re in a smart hospital?
“They have to ask patients or other physicians for information far less often. The data is electronically available, and ideally, not spread across different applications. Instead, it’s directly accessible at their usual workstation or mobile device.”
How would you define a smart hospital?
“Many think it’s about the level of digitization. But in fact, it’s about process quality. Digital tools are essential to achieving digital maturity, but they don’t make a hospital smart by themselves. The processes must be aligned. The staff must be able to operate the technology effectively, and the systems need to support workflows that make sense for the patient. I could implement a system where patients enter data at terminals, which might be digital but not smart, as they would be constantly shuffled around. That’s not good process design.”
The Charité University Hospital ranks eighth globally in the Newsweek and Statista ranking of the world’s smartest hospitals 2025. What innovative solutions are in use at Charité?
“We’ve optimized the admission process at dedicated, professional intake points because the way you admit patients significantly impacts the rest of their care. Also, all three of our campuses use the same system for patient data. If a patient returns or is transferred, the data is already there and goes with them. This ensures high data quality. Time-sensitive processes, such as lab and radiology orders, are completed electronically, and results are returned to the same system, saving a significant amount of time. Our medication management includes an electronic chart, which ensures visibility, legibility, and automatic interaction checks. These are only a few examples.”
Has AI already arrived in Charité’s life, or is it still science fiction?
“It depends on how we define AI. If we count traditional algorithms, they’ve been used for nearly 20 years, like ECGs that alert clinicians to potential critical issues. Structured data fields like radiology and lab systems already benefit from AI. However, the newer hype around large language models and generative AI is still experimental. Some small processes are already being supported, but we’re still gathering experience.”
How many AI projects are currently underway at Charité?
“We officially count about 70 to 80 projects. There are probably many more smaller experiments we’re not fully tracking. That’s why we’re cautious, not just to ensure progress but also to prevent errors. For example, we don’t want physicians uploading patient records to public GPT systems for help with documentation. We’re testing other tools that are GDPR-compliant and secure. We also have the Charité Lab for AI in Medicine, which defines legal frameworks and testing criteria for safe AI use. Plus, we’re part of the EU-funded TEF-Health project, which is building testing and certification centers for AI in medicine.”
How do you avoid getting stuck in “pilotitis”, the endless cycle of pilot projects?
“You can’t avoid it entirely, but that’s the point of pilots – to see what’s worthwhile. We’ve established a centralized project management office for projects that require significant resources. These are evaluated and prioritized. We ensure they meet data protection, cybersecurity, and medical safety as well as ethical standards. Of course, someone might still test something on their own with a smartphone, but critical projects follow a proper approval process. Most developers need access to clinical system data, which is tightly controlled.”
“For example, we piloted an AI system for acute kidney injury using historical data. Once it proved effective, we rolled it out. When a system like that is adopted, it becomes the standard for that use case. What we want to avoid is every department building its own AI systems, because inconsistent systems analyzing the same data can be dangerous.”
Even if generative AI isn’t fully deployed yet, what’s its potential in medicine?
“Its ability to organize and summarize complex information is a game-changer. In a fragmented healthcare system, generative AI can help by summarizing medical histories and supporting research. It’s already helpful in summarizing anamnesis conversations and extracting insights from vast data sets.”
In the U.S., major hospital information systems can already transcribe doctor-patient conversations directly into the electronic health record (EHR). Why isn’t this common in Europe?
“Our systems are much more fragmented. Integrating structured data into various EHR systems is complex. We’re testing ambient AI solutions, but usually, they produce a document that gets filed. Structuring that data remains a big challenge. As interoperability and structured data standards improve, it’ll become easier.”
Is interacting with the EHR via voice, instead of typing, the future of making notes?
“I’m sure about it. The original dream was a paperless hospital. Now the goal is a keyboard-less one. Ideally, not just doctors, but patients too, should interact with their EHRs. A patient should be able to ask their EHR monthly, “Is there anything I should know?” If the system says, “Your weight and blood pressure are trending in a dangerous direction,” and prompts them to see a doctor, that’s the kind of proactive care we need.”
“Currently, ambient AI systems mostly give us back a document to store. It’s useful, but we still need structured data if we want to make these systems truly intelligent.”
However, AI requires access to more than just medical data; it also needs lifestyle data, such as information from the ecosystem of life, home, or surroundings. Will the EHR eventually connect with smartwatches?
“I believe so. Although not directly, a smartphone could connect to both EHR data and insurer records and also track behaviors such as steps or sun exposure. Even shopping habits can reveal early warning signs. The British found that people buying fiber supplements or laxatives might be experiencing digestive issues as an early indicator for some diseases before seeking care. These insights are powerful.”
Be back tomorrow for the second part of our interview with Dr. Peter Gocke.
