Newsroom

The Execution Gap Is a Context Problem

Earnix published findings this week from a survey of more than 400 global insurance executives, including 40 UK insurance leaders.

Newsroom 15 June 2026

Earnix published findings this week from a survey of more than 400 global insurance executives, including 40 UK insurance leaders. The headline: more than half of UK insurers have integrated AI into core business functions, but the majority are now confronting what Earnix terms an execution gap > the distance between isolated AI pilots and consistent, enterprise-scale operational decision-making.

The numbers

The numbers are instructive in aggregate even if the sample is modest. Nearly all UK insurers are either already using or planning to use generative AI to process unstructured data. Investment in third-party data is increasing across 91% of respondents. And yet 30% acknowledge they are falling significantly behind customer expectations for personalisation, with operational processes struggling to keep pace with where the market is moving.

The Earnix commentary frames this as a workflow and embedding challenge. That is partly right. But it misses the deeper constraint, which is architectural.

The inputs are curated, the scope is defined

AI works well in a pilot because a pilot is contained. The inputs are curated, the scope is defined, and the consequences of a poor output are limited. Enterprise-scale AI decision-making is structurally different. In live underwriting, pricing and claims environments, an AI system is not operating on a single clean input. It is operating in a context, a web of prior decisions, policy history, claims relationships, risk signals and regulatory constraints that accumulates over time and varies by entity.

Most insurers have no infrastructure capable of holding that context in a form that AI can reliably use. They have data warehouses, policy administration systems, and CRMs that were not built for the purpose. When AI is bolted on top of fragmented data infrastructure, the execution gap is the inevitable result. The model is capable. The context layer beneath it is not.

Context is king

Closing that gap requires more than better tooling or faster deployment cycles. It requires an infrastructure decision: a persistent, structured context layer that carries the right information into every decision, maintains it across sessions, and can demonstrate after the fact what the system knew and on what basis it acted. That is what enterprise-scale AI actually runs on. Until insurers have it, the gap between ambition and execution will remain exactly where it is.