Insurance Claims Liability Optimisation

Context

A major European insurance group was processing high volumes of claims across multiple product lines. Every incoming claim required two judgement calls before meaningful work could begin: which department should handle it, and how liability should be apportioned between the insurer and the claimant. Both decisions were made manually, by individual assessors drawing on their own experience and knowledge of the policy terms.

The Challenge

Without a consistent framework for either decision, outcomes varied significantly from one assessor to the next. Claims regularly landed in the wrong processing teams, creating delays and internal handoffs that lengthened resolution times. Liability determinations — whether responsibility rested with the insurer, the customer, or some split of both — were applied inconsistently, leaving customers with decisions that felt arbitrary or unexplained.

The downstream effect was a persistent dispute rate that consumed operational capacity, eroded trust, and was driving measurable customer churn among policyholders who might otherwise have renewed.

Our Approach

We built an AI system that evaluates every incoming claim and produces two outputs: a routing recommendation directing the claim to the correct processing team, and a probabilistic liability assessment — quantifying the likely apportionment of responsibility between the insurer and the customer, expressed as a percentage split rather than a binary decision.

Both outputs are derived consistently from the claim submission itself, replacing assessor-by-assessor variation with a standardised, auditable process. Customers receive decisions grounded in the same logic regardless of who handles their case — making outcomes easier to explain and harder to dispute.

Results