Fragmented Data Limits AI Returns for Insurance Firms

Global insurance providers are aggressively scaling up their capital investments in artificial intelligence (AI) tools, aiming to drive down claims settlement costs and enhance underwriting accuracy. Despite these growing financial commitments, a substantial portion of the industry is still struggling to generate meaningful economic returns. Market data indicates that this lack of profitability stems primarily from obsolete legacy software architecture and poorly coordinated corporate data systems.

Investment Trajectories and Financial Projections

A research report published in March by Boston Consulting Group, Inc. (BCG) indicates that AI spending within the property and casualty (P&C) insurance sector is on a steep upward trajectory. Investment levels are projected to more than triple, reaching 1.9% of total corporate revenue in 2026, compared to trailing baseline figures.

The management consultancy’s data demonstrates that the eventual financial yield of AI depends heavily on how deeply it is integrated across an organisation. Insurers that successfully embed AI throughout their comprehensive operations stand to achieve an estimated 20% reduction in overall costs, alongside a potential increase in gross written premiums of up to 5%.

Isolated Deployments and Performance Statistics

While many insurers currently utilise AI within underwriting and claims processing to automate routine administrative tasks, refine pricing models, and identify fraudulent activity early, the deployment remains largely fragmented. BCG observed that these tools are frequently confined to isolated business units, a factor that severely limits their aggregate economic impact across the wider business.

Despite these structural hurdles, data gathered from current users reveals a strong appetite for the technology, alongside mixed operational outcomes. The table below details the performance metrics and future outlooks reported across the global insurance landscape:

Infrastructure Impediments and Expert Perspectives

A separate study published in April by A.M. Best Company, Inc. focused on the specific structural liabilities that hinder effective technological transition. Analysts pointed out that modern algorithmic data models are fundamentally incompatible with historical corporate IT infrastructure.

Kaitlin Piasecki, an industry research analyst at A.M. Best, stated within the publication:

“Legacy systems can create significant barriers when implementing AI because they simply were not built for this type of data integration.”

This perspective was reinforced by Sridhar Manyem, Senior Director of Industry Research and Analytics at A.M. Best, who noted that AI tools yield unreliable and sub-optimal outcomes when corporate data is fragmented or suffers from deficient governance protocols. Nonetheless, industry determination to adopt the technology remains intact, with two-thirds of surveyed insurers intending to expand their capital commitments to AI between 2026 and 2028 to overcome these legacy hurdles.