.png)
Last week, Thryve was in two places at once. Our Co-Founder and CEO Friedrich Lämmel and Chief Strategy Officer Nick Bosscher were in Singapore for the Health Insurance Innovation Congress Asia Pacific 2026, presenting to Asia's leading health and life insurers. Meanwhile, Co-Founder Paul Burggraf was in Berlin at AIDAQ 2026, giving a keynote on what actually creates durable advantage in healthcare AI.
The events were different in audience and geography, but the idea running through both was the same: the data layer is where it gets decided.
The theme of HIIC 2026 was From Cost Crisis to Care Innovation, and Friedrich opened with a question the room sat with: when your members compare you to the insurer next door, what are they actually comparing?
The Swiss Re Institute data behind the talk covered 12,101 consumers across twelve Asian markets. The headline finding, that 40% of people in advanced markets cite cost as their reason for not buying health insurance, gets cited constantly. What gets cited less is what sits directly underneath it in the same dataset. 46% of respondents said what they actually want is help staying well, in the form of health check-ups and screenings. That came in ahead of bigger payouts.
Financial anxiety around medical bills kept growing across nearly every country in the study, even as out-of-pocket costs fell as a share of total spending in most of the same markets. The system improved on paper, but nobody felt it. That gap between what the numbers say and what members experience is where the real opportunity sits.
The industry objection Friedrich addressed head-on is one most insurers know well: we do not invest in prevention because members churn. The argument sounds pragmatic. The data from insurers who moved anyway tells a different story. Personalization builds trust. Trust creates retention. Retention is what makes prevention economically viable. The logic runs in the direction most insurers have not tried.
Friedrich walked through Thryve's framework for the full member lifecycle: better awareness for well members, better self-care for those showing early risk signals, and better clinical care for those who are already unwell. He showed how wearable data and population-level intelligence connect each stage. For insurers willing to make the shift, the numbers are significant: risk prediction sensitivity can improve by up to 2.5x, long-term claims risk can be reduced by tens of millions per screened population, and operating profitability can improve by more than 10% through earlier interventions and better member targeting.
Nick and Alexander Spalding, our Insurance Consulting Lead, were on the ground throughout both days for conversations with leaders from Munich Re, Zurich, Allianz, Swiss Re, and others. The appetite in the region for this kind of infrastructure is real, and the conversations in Singapore reflected that.
On the same week, Paul was in Berlin making a related but distinct argument. His keynote at AIDAQ 2026 was titled From Foundation Models to Foundation Data, and it started with a provocation: look at the current landscape of healthcare AI companies. How many of them are actually companies? How many are features that the next foundation model release will absorb?
It is a question the sector has not fully answered yet. In 2024, summarizing a medical record was a product. In 2026, it is increasingly a model capability. Extracting structured information from clinical text followed the same arc. The pattern is clear enough that the more interesting question is not what AI can do with medical knowledge (guidelines, drug interactions, disease pathways, literature) but what it cannot do without something that models alone cannot provide.
AI does not know what happened to a specific patient last week. It does not know which medication they actually take versus which one was prescribed. It cannot see whether their activity dropped in the days before a symptom appeared, or whether a treatment actually changed their outcome. Medical intelligence without patient context is generic intelligence. It can tell you what the guideline says. It cannot tell you how far a real patient's care diverged from it, and what that divergence cost.
Paul's argument is that what remains scarce, and what creates durable advantage in healthcare AI, is not the model. Models are becoming abundant. What stays scarce is longitudinal patient data, continuous consumer health signals, real-world outcomes, and the rights and governance to use them. The organizations building that data layer now are building something that cannot be replicated by switching to a better model next year.
Paul showed Thryve's Population Level Intelligence platform as a live example of what this looks like in practice: the gap between what clinical guidelines prescribe and what actually happened to a patient over time, made visible and actionable for insurers and care teams.
The two conferences had different audiences asking different questions. HIIC was about what health insurers should become. AIDAQ was about where healthcare AI is heading. But the thread between them is the same.
The insurers who build a real relationship with their members, not just at claims time, but continuously, are building the data layer that makes everything else possible. Personalized engagement, earlier intervention, better risk prediction, AI that knows the patient and not just the population. That is not a technology problem. It is an infrastructure decision. And it is one that is being made right now.
If you are building a health or insurance product on wearable data and want to get to market faster with a ready-made data and intelligence layer, book a demo with Thryve!
