Insurers eye AI-led claims as bias & risk fears grow

Senior insurance professionals across the UK and Europe worry that artificial intelligence could lead to biased or unfair outcomes, even as most expect claims administration to become fully AI-managed within two years, according to research from embedded insurance provider EIP.

A survey of 250 senior professionals found that 87% are concerned about bias in AI-driven processes. At the same time, 90% expect claims administration to be managed end-to-end by AI within the next 24 months.

The results highlight a tension between confidence in automation's operational value and concern about the explainability and governance of decisions that affect customers. Claims processing and underwriting face close regulatory scrutiny in many European markets, increasing pressure on insurers and suppliers to show how automated decisions are made and to evidence consistent treatment.

Wider risk concerns

Bias was the most cited issue, but respondents also pointed to other risks. Data security and privacy ranked next at 23%. Regulatory non-compliance and system reliability and errors each registered at 21%, while job displacement or staff resistance came in at 20%.

Regulatory concerns were more pronounced among larger firms. One-third of respondents at bigger organisations cited regulatory non-compliance as a key concern.

Uneven comfort levels

Attitudes varied by use case. Respondents felt least comfortable automating claims submissions, with 40% selecting it as the area where they are least at ease. Underwriting recommendations followed at 39%, with customer interactions next at 35%.

The results suggest insurers distinguish between back-office tasks and front-line decisions that customers experience directly. They also point to heightened risk where automated systems touch the earliest stages of a claim, when information quality and fraud checks often shape the direction of a case.

Despite reservations about specific tasks, respondents expressed broad confidence that AI will reshape the sector. All said they expect AI to have a sizeable impact on insurance. More than half, 58%, expect major efficiency gains, while 33% anticipate it will transform the industry entirely.

Where AI is used

Insurers appear to be experimenting across several functions rather than focusing on a single area. Almost half of respondents, 45%, said they are already using or exploring AI for customer service and chatbots. Risk modelling and data analytics followed at 43%.

Claims management was cited by 41% of respondents, with underwriting and pricing and fraud detection both at 40%. Marketing came in at 38%.

Human involvement also remains a near-universal expectation. Some 99% of respondents said AI-driven outcomes should have some level of human oversight.

Transparency expectations

Respondents highlighted transparency measures as a way to address concerns. The most frequently cited reassurance was the use of transparent algorithms and decision logs, selected by 39%.

Integration and vendor support also featured strongly in buying criteria. Some 27% said simple integration is the most important factor when considering an AI tool, while the same share highlighted partnership and support.

Cost appeared less influential. Only 10% said cost and return on investment would strongly influence their assessment of an AI solution.

Ross Sinclair, founder and CEO of EIP, linked the results to regulation and the need for audit trails in claims decisions.

"Insurance is, by necessity, a heavily regulated industry, which makes introducing AI complicated. This research shows that the development of industry-specific capabilities could help to overcome challenges and accelerate the adoption of AI right across the industry, whether that's for customer service, risk modelling or even claims management and decisioning," said Ross Sinclair, founder and CEO of EIP.

Sinclair said claims decisions must be explainable and auditable. He also argued for rules-based approaches that insurers can configure to policy terms and compliance requirements.

David Mitchell-Dawson, EIP's director of product, said bias concerns pre-date the current wave of AI tools.

"It is important to remember that bias is an issue for insurers regardless of whether they are using AI or not. It has always been a challenge for the industry. A rules-based engine addresses this because it is not subject to any inherent biases," said Mitchell-Dawson.

He also pointed to a faster pace of adoption than is often associated with insurance. "Insurers told us that they're expecting their full end-to-end claims administration to be handled by AI within the next 15 months on average, so frankly if you're not doing anything now, you're already too late," he said.

EIP, founded in 2004, provides embedded insurance programmes and software for insurers and partners in sectors including banking, retail and telecommunications. It supports more than 2 million end customers, manages £220 million in annual premiums, and processes more than 300,000 claims each year.

The research was conducted by Censuswide among senior insurance professionals in the UK and across Europe, including Italy, Germany and Spain. It focused on attitudes to AI adoption, governance and operational change.