Most AI efforts focus on improving the model. In practice, the success of a medical AI product depends on something else: what you check before and around the model.

This session presents a practical checklist for building AI-powered medical products. It covers key questions around real-world data, robustness, workflow integration, safety, and business viability.

The key takeaway is that the “secret sauce” is not full automation. It is stepping back and designing a system where AI supports humans at the right moments; combining the efficiency of automation with human judgment, oversight, and accountability. Ultimately, value is not created by the model alone, but by how it improves the workflow in which it operates.

Dr. Ágoston Török is Head of AI Technology at Straumann Group, where he leads the development of AI systems for automated imaging analysis and treatment planning.

With a background in cognitive neuroscience and over a decade in applied AI, he has worked across startups and scale-ups, contributing to ventures with successful exits and helping secure multi-million-euro investments.

His work focuses on building AI-powered products in regulated domains that impact clinical workflows and have affected the lives and well-being of millions of users. He is particularly interested in turning machine learning models into practical, scalable systems that work in the real world.