
We all use healthcare. Some of us more than others, but who hasn’t been to see a doctor this year? Health is a part of our everyday lives, and it’s not immune to the rise of AI.
The pro-AI lobby is always banging on that it can help cure cancer. Sure, maybe it will, but AI is likely to have a much more integral impact on your everyday healthcare. From detecting strokes and tumours to predicting infections in newborn babies and helping with paperwork, AI is moving into hospitals with sirens blaring. AI saves lives. But it’s also creeping into your home. AI doesn’t get tired. It doesn’t need coffee. It’s there day and night to answer any health question you can think of (and according to a Bitkom Research survey from 2025, 45% of people in Germany already do).
Doctors have raised serious concerns. AI is designed to be helpful and responsive to the user, which can mean telling people what they want to hear (or what AI thinks you want to hear). But a good doctor confronts a patient. They ask those awkward and uncomfortable questions, get their hands dirty and point out the things you’d rather not hear. The good ones don’t enable you. That human-to-human encounter can be an essential part of medical practice. AI can simulate sympathy, but it doesn’t feel empathy. It can respond with warmth and understanding, but it has neither. So why should we trust it with our lives?
AI at Charité

Berlin is one of the centres for research in medicine. In November 2025, Charité and the Deutsches Herzzentrum established the Institute for Artificial Intelligence in Medicine (IKIM), led by Professor Alexander Meyer. The institute’s goal is to develop AI applications and bring them into real clinical practice. Meyer is frank about the challenge. He admits that many AI systems fail to make the transition into clinical practice. “They often don’t reach their goal,” he stated in an official press release, “and not because of poor algorithms, but because of a lack of structures, lack of evidence and lack of trust.”
One of Berlin’s strongest areas of research is in medical imaging. Modern medical images, such as MRI and CT scans, contain enormous amounts of information. AI can be trained to recognise patterns that would be difficult or time-consuming for a human to identify. At Charité’s Lab for Artificial Intelligence in Medicine, known as CLAIM, researchers are developing AI systems to predict conditions including strokes. Their systems can analyse brain images and help identify abnormalities and important structures. The results earned the CLAIM team first place in two international MICCAI challenges involving medical image analysis. The concept is simple: a doctor looks at a scan, while the computer looks at the same scan from a different perspective. So the AI doesn’t replace the doctor; it provides another set of eyes. Another set of eyes could be literally lifesaving in an emergency, when every second counts.

Another extraordinary field of research is in cancer. Specialists at Charité and the Berlin Institute of Health developed an AI model called crossNN, which analyses epigenetic information from tumours. Rather than simply looking at the shape of a tumour, the system examines biological information and compares it with a large reference database. In the researchers’ tests, the model correctly classified brain tumours in 99.1% of cases. It could also distinguish more than 170 tumour types, achieving an accuracy of 97.8% in the reported testing. Comparing a patient’s tumour with thousands of other tumour profiles would be a near-impossible task for a human.
However, these achievements are yet to prove themselves in a real hospital environment. In a controlled research environment, the algorithm works well, but it can still encounter unexpected problems when introduced to the messy world of real life.
Explaining, Preventing, Paperwork
One thing AI struggles with is explaining why doctors make a decision. It can make the comparisons, sure, but lacks the critical thinking of a trained medical professional. Researchers at Charité and Technische Universität Berlin (TU) are therefore working on ‘explainable AI’. Some systems can indicate which parts of a medical image might influence a doctor’s decision, for example by producing a heat map over a tissue sample. That way, doctors don’t blindly trust a machine. They can question its recommendations, compare it with their own judgement and understand when the system might be making a mistake. The goal is not to create an intelligent machine that does the work, but to create an intelligent machine that humans can safely work alongside.
Perhaps one of the most unusual Berlin projects in AI medicine is called nAIonate. The project is rather new, and is scheduled to run from 2026 to 2028, and investigates whether AI can identify possible outbreaks of bacterial infections in neonatal medicine at an early stage. The project involves approximately 1,800 neonatal patients and analysing routine clinical data, microbiological results and genetic information. If an outbreak can be recognised earlier, infection-control specialists may be able to react before more patients become affected. This is one of the most exciting promises of AI in medicine: the move from treating serious diseases to predicting and possibly preventing them.
It’s not trying to replace the doctor; it’s rather an always-on assistant.
A much simpler use of AI is delegating paper work. Doctors spend an enormous amount of time writing notes, reports and discharge letters. While it may not sound revolutionary, AI could take over some of this administrative work. Charité researchers are investigating AI-supported documentation and large-language-model applications to summarise medical information, check documents for inconsistencies or help doctors navigate clinical guidelines.
Berlin’s wider AI startup scene is also moving into healthcare with some bringing AI directly to the patient. Berlin startup Big Picture has developed Adele, a voice-based health assistant that reminds its users to take medication, hydrate, monitor vital signs, detect falls and help trigger an emergency call. It’s a kind of digital care companion, particularly for people who want to remain independent despite aging or chronic illness. “AI can provide a first line of support, to have something that helps you evaluate whether you should go to the ER,” says Big Picture CEO Maximilian Vogel. “It’s not trying to replace the doctor; it’s rather an always-on assistant.” With a health care system under strain, Big Picture claims that AI has the potential to lighten the load. Vogel thinks that in the next two to three years, AI will fill these practical assistant roles, rather than replacing doctors.

We All Make Mistakes
So, could AI really save your life? Quite simply, yes. It could. An AI system might identify a stroke, recognise a suspicious tumour, detect the early signs of infection or warn doctors that a patient is deteriorating. But humans and AI are more similar than we like to think. Namely, we both make mistakes.
One of the biggest concerns is bias. AI learns from data. If the data used to train an algorithm do not represent different patient groups equally, the system may work better for some people than others. Then there’s privacy. Medical records contain some of the most sensitive information imaginable. Researchers need large datasets to develop AI, but patients also have a right to control and protect their personal information. That’s not to mention the problem of responsibility. If a doctor ignores an AI recommendation and the patient dies, who’s responsible? And if the doctor follows an AI recommendation that turns out to be wrong? Then what?
AI is better at analysing millions of datasets at one time than a doctor. A doctor is better at understanding an individual patient, communicating difficult information, considering personal circumstances and taking responsibility for decisions than an AI agent. For now at least, the human element feels impossible to automate. Some of the world’s top researchers right here in Berlin are attempting to find the balance. They’re already proven what AI can do. Now it’s about proving that these technologies can be used safely and reliably in everyday healthcare.
