AI in European Insurance: Applications, Risks and Regulation
From Allianz to Baloise: four case studies and three regulators show that AI adoption in insurance is a governance story, not only a technology one.
The big picture from supervisors' reports
In this independent research I analysed AI and Generative AI adoption in the European insurance industry, focusing on Germany and Switzerland. The basis was official publications and surveys by EIOPA, BaFin and FINMA — sources that give a quantitative picture: AI use in underwriting, claims handling, fraud detection and customer service is expanding, but at different speeds across large and small firms.
Four case studies
The Allianz, ERGO, Zurich and Baloise case studies revealed three shared patterns: first, generative models are used more for internal productivity than customer products; second, data governance and model explainability precede entry into sensitive processes; third, each firm runs a different blend of technology and business teams to monitor models.
Risks and the regulatory approach
Risks fall into three groups: model risk (bias and error in underwriting and claims decisions), data risk (privacy and quality) and operational risk (dependence on external providers). Supervisors respond with risk-based regulation: from the EU AI framework to BaFin guidance and FINMA's supervisory approach — all emphasising explainability, human accountability and continuous monitoring.
My conclusion for emerging markets is clear: the European experience says the right order is “governance first, scale second”; firms that built governance frameworks before scaling applications experienced less risk and faster adoption in sensitive processes.
Key takeaways
- AI adoption in insurance is first a governance problem, then a technology one.
- European supervisors agree on three axes: explainability, human accountability, continuous monitoring.
- For emerging markets: governance first, scale second.