At the Reinsurance Rendezvous Zurich, organized by the Swiss Insurtech Hub in collaboration with the Boston Consulting Group, leading figures from the Swiss reinsurance industry gathered in Zurich on June 9 for an open discussion on the question currently most pressing for the industry: Where does artificial intelligence actually deliver business results?
The evening began with an opening address by Coralie Ming, a principal at the Boston Consulting Group (BCG), who shared two key figures: $4.7 billion in global insurtech funding in 2025—marking the first growth since 2021—and a 44 percent share of AI-driven deals. Despite this momentum, BCG also highlighted the sobering reality: only 30 percent of companies report actually deriving value from AI. The consultants cited technology-driven initiatives without a clear business anchor, piecemeal solutions lacking end-to-end thinking, and an underestimation of the so-called 70 percent factor—that portion of change work in organization, processes, and people that determines success or failure—as the main reasons.
Where value is truly created
During the subsequent panel discussion, moderated by Raphael Troitzsch (BCG), it quickly became clear that the theoretical framework is indeed being put into practice. Sofia Kyriakopoulou, Chief Technology, Data, and AI Officer at SCOR, described how her company has introduced AI assistants for medical underwriting in the life and health insurance sector. The system summarizes medical reports, extracts relevant information, and prepares the basis for decision-making. The underwriters’ initial reaction? Caution. Employees manually verified the outputs, which initially even reduced productivity. It wasn’t until about three months later that trust began to grow; after six months, the J-curve became apparent: 30 to 40 percent greater efficiency, a more standardized understanding of what constitutes good work, and a significantly faster onboarding process for new hires. In the commercial sector, SCOR reports for the first time that processes that used to take days now take minutes.
Beat Kramer, Group AI Enablement Lead at Swiss Re, highlighted two particularly valuable areas of application: claims management and growth areas. In the property and casualty sector, the value contribution is directly measurable, as a claim that is avoided has an immediate impact on the bottom line. In growth areas, however, AI increases capacity without the need to build up additional resources. The most difficult question—whether AI actually leads to better underwriting decisions—remains unanswered for now. Proxy metrics can be measured, such as the conversion rate from submissions to deals, but conclusive proof still requires conviction, not just data.
AI is making its way into the real world
Michael Unger of NVIDIA broadened the perspective. He described the evolution of AI in four phases: prescriptive AI for classification, generative AI in the vein of ChatGPT, agent-based AI with self-organizing systems, and now physical and industrial AI, which encompasses robotics, autonomous driving, and production simulations. This physical world accounts for the majority of global GDP. The current AI transformation has so far affected only a small segment of the digital economy; what comes next will be many times larger.
The rapid expansion of data centers is of direct relevance to insurers and reinsurers. According to industry estimates, data centers alone could generate $20 billion in insurance and reinsurance premiums by 2030. Individual data centers today have construction values of up to $20 billion, an amount that far exceeds an insurer’s typical single-risk program. Added to this is the geographical concentration: many of these facilities are being built in U.S. regions with increased exposure to tornadoes and hail. The issue of accumulation poses a new and significant challenge for the industry.
The Talent Question and the Future of Job Roles
All panelists agreed that AI is profoundly changing job roles, but not necessarily eliminating them. Beat Kramer described a possible reversal: Whereas claims adjusters currently spend 80 percent of their time on analysis, that ratio could shift: 20 percent analysis, 80 percent action. The real challenge is knowledge transfer: If AI takes over the basic analytical work, junior staff will lack precisely those learning opportunities through which expert knowledge has traditionally been passed on.
Sofia Kyriakopoulou advocated actively documenting the thought processes of experienced underwriters and making them available for training the next generation, whether human or machine. She observed that AI is increasingly becoming an extension of human expertise. Those who are good at making judgments will become even better with AI; those who work superficially will remain so even with AI.
Three insurtech companies show where AI is making a tangible impact today
During the event, three members of the Swiss Insurtech Hub introduced themselves and demonstrated that the path from AI theory to practice is shorter than is often assumed.
Parashift, a startup based in the Basel region, specializes in AI-powered document processing—a seemingly unspectacular task that is, however, highly relevant in the insurance industry. Claims, onboarding, policy entry: according to Parashift, up to 95 percent of these processes can be automated at twice the speed. In doing so, the company relies on the principle of data sovereignty. The data remains in Switzerland, Germany, or within the EU and is limited to what is internally referred to as “Trusted Data”: AI outputs that are secured by integrated guardrails so that employees can trust them without having to manually verify every result. The process typically begins with a one- to two-day proof of concept using historical customer data; according to Parashift, a return on investment can be demonstrated within just a few months.
ClearSpeed, on the other hand, takes a completely different approach: Originally developed for military special forces, the company analyzes risks based on the human voice—not using AI, but rather neuroscience. Two to four simple yes-or-no questions are enough to provide a real-time risk assessment, regardless of language, dialect, or cultural background. The technology detects unconscious neural signals that arise when someone gives an answer that contradicts their inner experience. This is not a lie detector, but a precise early-stage filter, comparable to a metal detector at an airport. With a target accuracy of over 98 percent, ClearSpeed claims to be far more accurate than polygraph tests. In the insurance industry, the solution is used for claims management, fraud detection, and underwriting, with the added benefit that processing times for non-problematic cases are drastically reduced and customer satisfaction increases.
Finally, MavenBlue, based in the Netherlands, addresses a problem that is all too familiar in insurance and reinsurance circles: the slowness of balance sheet analysis. Born out of the concrete frustration of a former Chief Investment Officer at Fortis International, who had to wait months for answers to regulatory and strategic balance sheet questions, the Dutch insurtech offers a GPU-based solution for fast, iterative balance sheet projection. Whether it’s deterministic ORSA scenarios, stochastic modeling, the optimization of reinsurance programs, or the calculation of internal rates of return for private equity-backed companies: MavenBlue aims to transform a process that used to take six months into one that delivers results in hours. After nine years of market penetration in the Netherlands—with 85 percent market coverage, according to the company—it is now actively expanding into new markets and explicitly seeking to engage with the reinsurance industry
Zurich as a Resonance Chamber
What became clear during the evening at the Reinsurance Rendezvous Zurich is that the Swiss market possesses the expertise, the networks, and the determination not only to talk about AI, but to embed it at the core of its business. The real question is no longer whether, but how quickly—and with what level of human judgment behind it.
Binci Heeb
Read also: Embedded insurance in everyday life
Search:
Sponsors:
