tech11 · AI Insights
Ask an insurance executive whether their organisation uses AI, and the answer is increasingly “yes”. Ask whether they trust it enough to let it near a claim, and the conversation gets more careful. Insurance does not need a more impressive AI. It needs a more reliable one, and the gap between using AI and trusting it is where the real work lies.
AI is already part of European insurance. In its 2023 digitalisation survey, published in April 2024, EIOPA found that 50% of responding insurers already used AI in non-life insurance and 24% in life insurance, with many more planning to follow. Customer service, fraud detection and claims processing were among the most common uses.
So the debate has moved on, from “should we use AI?” to “how do we use it responsibly, and how do we show that we do?”
In insurance, a fast answer, or even a frequently correct one, is not the standard. The standard is a decision that can be defended: one the insurer can explain, show to be consistent, and support in front of a risk committee, an internal audit or, sometimes, a court.
That is because insurers manage more than data. They manage contracts, coverage terms and exclusions, expert evidence, timelines and responsibilities, and their decisions carry financial, reputational and often legal weight. Interpreting coverage, assessing damage, spotting fraud or handling a complex claim all need context, rules and reasoning, not only a statistically likely output. An AI that offers a probable answer can help with supporting tasks. Decisions that must be justified ask for more.
Trust does not come from a more powerful model. In our view, it rests on four things.
A clear role. An AI agent should have a defined job, defined tools and defined permissions, much like a colleague in a well-run team. When it is clear what the AI may and may not do, it becomes easier to rely on.
Human authority. AI can read, structure and prepare. People decide. Regulators name the risk of over-reliance explicitly: for high-risk systems, the EU AI Act asks that the people overseeing AI stay aware of the tendency to trust its output too readily, and that they can override or reverse it.
Traceability. In insurance, “what happened, and why?” must always have an answer. For high-risk systems, the EU AI Act requires automatic logging of events over the system’s lifetime. Even where a use case is not high-risk, good practice points the same way.
The insurer’s own knowledge. Good decisions rely on rules and expertise that often sit in scattered documents, systems and individual experience. AI that works with an insurer’s own business rules, rather than around them, helps decisions stay consistent from one case to the next.
In August 2025, EIOPA published an Opinion on AI governance and risk management. Its message is practical: insurance legislation already applies to all AI systems used in insurance, and governance should be proportionate to the risk of each use case. It stresses human oversight, clear responsibilities across functions such as compliance, actuarial and data protection, and sound data governance.
Not every insurance use of AI is classed as high-risk under the AI Act. Its Annex III, for example, names risk assessment and pricing for natural persons in life and health insurance. For high-risk systems, the Act goes beyond effectiveness and sets requirements such as record-keeping, human oversight, accuracy and robustness. The direction is the same everywhere: governance is not something added at the end.
No moment tests trust like a claim. A policyholder reports a loss, often on a bad day, and expects to be understood the first time.
One US survey of 800 policyholders, commissioned by a claims-software provider, illustrates the point. Among those who explained what happened only once, 86% were satisfied with their claims experience. Among those who had to repeat the details six or more times, only 7% were. The survey is American and vendor-commissioned, so read it as an indication, not a benchmark. The lesson still holds: being understood early shapes everything that follows.
In insurance, trust has never been cosmetic. Policyholders rely on their policy being interpreted correctly, insurers on their judgement standing up to scrutiny, supervisors on governance being in place. Increasingly, that trust will also depend on the kind of AI an insurer chooses to adopt.
At tech11, we build cloud-native core software for insurers. Our vision for AI follows one idea: AI that supports the work, not the authority. We are building towards specialised agents, each with a clear responsibility, that work inside real insurance processes in three steps:
AI prepares, analyses and suggests. People decide. The aim is not to make decisions less human, but to make them clearer, better supported and easier to stand behind.
This is a vision we are still building towards, and we would rather say so openly. Trust is earned step by step, and it starts with being honest about where we are.
If you work in claims, underwriting or insurance IT, we would like to hear from you: what would you want AI to do for your teams, and what would you never want it to do?
Talk to us: tech11.com/en/sales
Sources: EIOPA, Report on the digitalisation of the European insurance sector (April 2024); EIOPA, Opinion on AI governance and risk management (August 2025); Regulation (EU) 2024/1689 (AI Act), Chapter III Section 2 (including Articles 12 and 14) and Annex III; Hi Marley, Voice of the Policyholder survey on FNOL.