Two days in Berlin, one question kept coming back: if AI in healthcare is gaining momentum, whose data is it built on?
Arvato Systems’ 11th Life Science Symposium brought pharma, clinicians, researchers and founders together under the theme Momentum. Day 2, “Direction Before Disruption”, made the strongest case yet that the next leap will not come from better algorithms. It will come from data that reflects the people medicine is meant to serve.
The bottleneck is data, not models
Dr. Rafik Fellague-Chebra opened the day with “From Computation to Care”. Of every 100 US patients with advanced solid tumours, about 8 are eligible for a genomically targeted therapy and only 5 benefit. Two patients can share the same mutation, drug and dose and still have opposite outcomes, because sex, BMI, co-medications and even a recent course of antibiotics change how treatment works. His verdict on readiness: science, technology and regulation are moving. Operations and data are the constraint.
The panel on multimodal data with Dr. Jost Leemhuis, Dr. Nadine Diersch and Dr. Konstantin Thierbach picked up the same thread. Combining data is where clinical value appears, but only with access, governance frameworks and data that is truly AI-ready.
Who the data sees
If cleaner data is the answer, the She Shifts Systems focus session asked the harder question: clean for whom? Prof. Dr. Sylvia Thun (BIH, Charité) showed how women disappear from health data, in trials, symptoms, coding and analysis, and how the European Health Data Space can help close that gap. Ida Tin (Clue, SPRIND) looked at the next breakthrough in women’s health and why the market is ready, even if the data is not.
Our CEO and Co-Founder Dr. Manasi Aichmüller-Ratnaparkhe widened the lens from gender to populations. Much of medical knowledge was built on a narrow reference group, and AI now applies those assumptions at scale. She also reminded the room that diversity does not only exist between countries. It exists inside every healthcare system, including Germany’s.
“Every stage in Berlin arrived at the same point: the model is rarely the problem, the data is. When a patient is missing from the data, she is missing from the evidence, the trial and, in the end, the treatment. Building representative data is not a side project. It is how better medicine reaches everyone.” Dr. Manasi Aichmüller-Ratnaparkhe, CEO and Co-Founder, PAICON
The session closed with the panel “Invisible Patients: If Data Doesn’t See You, Healthcare Can’t Treat You”, bringing Sylvia Thun, Ida Tin, Birgit Bauer and Manasi together. Infrastructure, blind spots and patient voices kept returning as the three things that decide who is visible.
Diversity is a business case
The clearest shift in Berlin was in how the topic was framed. Representative data is no longer only an ethical goal. The EU AI Act now requires high-risk AI to be trained on sufficiently representative data, and pharma is looking for growth beyond the markets it already measures. As the graphic recording put it: it is not just a women’s issue, it is a business case. Success cannot mean financial return alone, and the opportunity needs funding.
What we take home is simple. Momentum is real, but direction matters more than speed. The infrastructure that makes blind spots visible is what will decide which AI reaches the clinic, and who it works for.
Thank you to Arvato Systems for putting this conversation on the main stage, and to everyone who joined us. Learn more about our work and remaining84.
Note: Text drafted with the support of AI tools and verified by our editorial team.