How can we move from experimenting with AI to solutions that support key business processes? What role do AI agents play in areas such as underwriting or claims handling? How can the potential of AI be reconciled with ever-increasing regulatory requirements? These questions were answered by leaders in the insurance industry who have successfully moved from experimenting with AI to implementations delivering tangible business results.
Key takeaways from this article
- The greatest business value comes from a comprehensive transformation of end-to-end processes using collaborative AI agents.
- The successful implementation of agentic AI requires structured data, a consistent technological architecture and a clear division of responsibilities between specialists and AI agents.
- In the insurance industry, AI acts as an advanced assistant supporting experts, whilst responsibility for business and financial decisions remains with humans.
How industry leaders are putting AI into practice
The webinar ‘AI Agents in the Insurance Industry: From Experiments to Practical Implementation’, organised by Gazeta Ubezpieczeniowa and Britenet, was attended by Bartosz Dołkowski (IT Managing Director and CIO at STU | STUnŻ ERGO Hestia S.A.), Dorota Urbańska (Director of the Claims Department at LUX Med Ubezpieczenia), Marek Majdak (CTO at PZU) and Piotr Wegner (Solutions Director at Britenet). The discussion was moderated by Aleksandra E. Wysocka, Editor-in-Chief of Gazeta Ubezpieczeniowa.
In their presentations, industry experts demonstrated how their organisations are using agentic AI to streamline key business processes. They also shared their experiences of implementing AI in areas such as underwriting and claims handling and highlighted the main challenges involved in scaling these solutions. Here are some of the key insights from the event.
Insight 1: From individual implementations to process transformation
Many companies have already completed their first AI implementations, but the greatest business value only emerges when AI moves beyond simply streamlining individual tasks and becomes part of a broader process transformation.
Bartosz Dołkowski of ERGO Hestia pointed out that the future lies in end-to-end processes, in which AI agents working together support complete business workflows. This approach ensures the continuity of information between successive stages of the process, enables better use of data, and minimises the loss of context between systems.
In practice, this means moving away from piecemeal optimisation of individual steps towards building a cohesive environment in which data, processes and technology function as a single ecosystem.
Insight 2: Process orchestration rather than the automation of individual tasks
Rising customer expectations and the increasing complexity of processes mean that the traditional approach to automation is no longer sufficient – organisations must design a new model of collaboration between people, AI and systems.
As Dorota Urbańska emphasised, the future of the insurance market will not be based on the implementation of further AI tools, but on the effective orchestration of entire processes. In such a model, technology takes over administrative and analytical tasks, whilst humans focus on making decisions that require experience, interpretation and business accountability.
This approach is already being used at LUX MED Ubezpieczenia, where artificial intelligence supports claims handling processes but does not act as an independent decision-maker. The organisation has adopted a model of collaboration between people, AI agents and RPA (Robotic Process Automation) bots, in which technology supports the analysis and execution of tasks, whilst responsibility for decisions remains with humans.
Insight 3: The new role of AI in underwriting
The greatest challenge associated with the use of AI in underwriting processes is not the decision-making itself, but rather the preparation of the comprehensive set of information required to make those decisions. AI delivers the highest value in this area by reducing the time taken for analysis and allowing experts to focus on risk assessment.
Marek Majdak presented an example of an implementation at PZU, where AI agents assist with the analysis of medical documentation, providing underwriters with a structured set of information necessary for risk assessment.
Experts unanimously emphasised that, under the current regulatory framework, artificial intelligence does not replace the underwriter – its role is to provide recommendations, analyses and structured context to support the decision-making process.
Insight 4: Data as the foundation of agentic AI
Many AI projects fail to deliver the expected results. These failures are not due to limitations inherent in the AI models themselves, but rather to issues with data quality, data availability and inconsistent information architecture.
During the webinar, it was emphasised on several occasions that the effectiveness of agentic AI depends primarily on the quality of the data and the information architecture. Piotr Wegner highlighted three elements essential for building scalable agentic AI solutions:
- consistent data semantics and a uniform understanding of business concepts,
- access to up-to-date information and the ability to process it in near real time,
- full auditability of processes and actions undertaken by AI agents.
Unless these conditions are met, even the most advanced AI models will not be able to operate in a predictable and safe manner.
Insight 5: Humans remain a key part of the process
Artificial intelligence is becoming increasingly effective in supporting analytical and operational processes; however, responsibility for business, financial, and regulatory decisions must remain with humans.
In practice, this means establishing a collaborative model in which AI analyses data, identifies patterns, and generates recommendations, while experts make decisions that require experience, industry knowledge, and risk assessment.
The webinar demonstrated that the successful implementation of agentic AI requires not only the right technology, but also well-organised data, properly designed processes and a clear division of responsibilities. If you’d like to learn from the practical experiences of insurance market leaders, watch the webinar recording or register for the next session to find out how to implement AI in the insurance industry in compliance with the AI Act, DORA and GDPR.
