Insurance deployed AI early and widely: claims triage, assistants on the contact floor, sales support across distribution networks, copilots at head office. Adoption is no longer the question. The question is whether that AI does what everyone believes it does. In this sector an agent error is paid later, in human rework and in delay for the policyholder.
Idun Group addresses the working AI already deployed inside the insurer and what it returns. We do not address actuarial models, pricing, risk selection or automated claims decisions. We analyse usage, never people, and never individual case files.
Claims handling, customer relations and contact floors, sales and distribution, head-office functions, and IT and engineering teams. On claims the analysis compares the routing the agent gives with what the handler then does with the file: that gap is the share of work the agent was meant to absorb and quietly hands back.
Industrialise where usage is broad and downstream rework is low. Consolidate where one use case is re-implemented across entities or products. Cut where rework exceeds the time returned. Each outcome maps to an immediate budget action.
We analyse what is genuinely sent to the agents, not only what they answer. That allows context hygiene to be set: less data served, answers just as good, reduced exposure. It is the question the risk function and compliance ask before any extension.
None of our publishable references is an insurer, and we would rather write that than let you assume otherwise. More than 40 AI assessments in under a year across more than 10 large enterprise clients. At Inskip, a consulting firm of around forty people, measuring real Claude usage established 85 % real activation, 33 users across 39 seats.
