Discussions about AI in the insurance sector have long followed a simple script – the technology promised faster underwriting or better customer service, yet all conversations inevitably drifted towards regulations, risk and security. During a webinar organised by Polish insurance journal Gazeta Ubezpieczeniowa in collaboration with Britenet, industry leaders with successful AI deployments under their belts discussed facts, outcomes and opportunities.
Key takeaways from this article
- AI Governance accelerates the adoption of artificial intelligence, but only when it is an integral part of the project from the very beginning.
- Regulatory compliance should be built into the solution right at the design stage.
- Organisations need AI Governance frameworks that extend beyond the AI Act, combining ISO 42001 standards with practical guidelines (e.g. NIST AI RMF).
- Human oversight remains essential, even in highly automated AI processes.
- The success of AI increasingly depends on organisational capabilities rather than mere access to technology.
Has AI Governance become the driving force of modern business?
One of the fundamental hurdles insurers have long struggled with is scaling AI across critical processes without losing control.
The moment artificial intelligence moved past the pilot phase and began influencing customer-facing areas, AI Governance ceased to be merely a legal issue. Conversely, companies that treat these procedures as an afterthought to be addressed post-deployment face severe challenges when attempting to scale their solutions further.
This was exactly one of the main topics industry leaders analysed during the AI Governance in Insurance webinar – how to prepare an organisation for the AI Act and DORA.
The AI Act alone is not enough to build an effective operating model
Damian Jagusz, Chief Information Security Officer at Ergo Hestia, argued that relying exclusively on the EU AI Act risks creating a compliance programme without an operating model. Instead, his organisation combines three complementary elements: the EU AI Act, the ISO 42001 standard and the NIST AI Risk Management Framework.
The AI Act tells us what we need to achieve, while ISO and NIST help us understand how to achieve it.
Standards and practical guidelines help organisations develop effective methods of operation. For insurers grappling with complex technological infrastructures and growing regulatory scrutiny, this layered approach is vastly more practical than treating legislation alone as a recipe for success.
Compliance by Design reduces risk before it appears
Tomasz Kasprzak, Experience & Innovation at Trasti, presented perhaps the clearest example of what compliance by design looks like in practice.
Bianca, Trasti’s AI-powered claims assistant, iis a solution that incorporates elements such as human oversight prior to claims decisions, transparent customer communication, complete audit trails, post-deployment monitoring and continuous validation were built in at the very initial stage of the project.
This example highlighted a pivotal shift in the mindset of insurance companies – instead of focusing on how to make an existing AI system compliant with regulations, compliance must be built into the system itself right at the design stage.
Human oversight remains a strategic requirement
Despite the rapid advancements in generative AI and agentic systems, none of the speakers suggested removing humans from the processes. Ergo Hestia's GPT-based assistant for insurance agents perfectly illustrates why.
With this solution, every response is verified against defined quality thresholds – weaker answers are automatically escalated, and samples of successful interactions still undergo manual review. As a result, the company maintains its response quality above its internal KPI of 90%.
Competitive advantage stems from organisational capabilities
Competitive advantage is no longer built simply through access to AI models, as these are becoming universally available. What truly distinguishes organisations is the ability to responsibly manage artificial intelligence, which opens the door to unlocking its full potential.
For executive teams, this means shifting investments away from chasing the latest models and towards building AI Governance, competencies, operational discipline and organisational capabilities.
Local AI can improve security, control and cost predictability
Piotr Wegner, Solutions Director at Britenet, introduced another aspect that is drawing increasing attention across the industry – local AI deployments.
Rather than relying solely on cloud services, some organisations are considering isolated AI environments that provide tighter control over sensitive data, more predictable operational costs, greater insight into model behaviour and better validation prior to production deployment.
At Britenet, we have created an isolated AI laboratory where models and AI agents are tested and benchmarked before gaining access to real business processes.
An AI agent should only gain access to data and real-world tasks once it can be proven that the organisation has full control over it. This principle extends beyond the infrastructure itself. It reflects a broader shift in approach, grounded in building trust before entrusting artificial intelligence with a wider scope of tasks.
AI Governance should be active even before scaling begins
One of the final conclusions of the webinar was that the most advanced organisations are not necessarily those experimenting with the newest models, but rather those investing in the discipline required to operate AI consistently, transparently and safely.
Conclusion
Competitive advantage increasingly depends less on artificial intelligence itself, and more on the organisational maturity required to utilise it properly.
Insurance industry leaders who have already deployed AI in production environments demonstrated that scaling artificial intelligence securely and profitably is not just a matter of modern technology, but above all, modernised data foundations, mature operating models and a robust AI Governance framework. If you want to explore the practical approach of these experts, watch the full recording of the event.
