Most Banks and Insurers Trapped at AI Pilot Stage as Enterprise Orchestration Proves Elusive, Capgemini Finds
Most Banks and Insurers Trapped at AI Pilot Stage as Enterprise Orchestration Proves Elusive, Capgemini Finds #
Most financial institutions have not moved multi-agent AI beyond isolated experiments, according to research published by the Capgemini Research Institute. The finding points to a gap between stated ambition and operational progress across the sector.
Capgemini’s World Cloud Report in Financial Services 2026, based on a survey of 1,100 financial services executives, found that 59% of banks and insurers are piloting or deploying multi-agent AI systems. Only 10% have reached enterprise-wide orchestration. About half of respondents said they cannot consistently demonstrate a return on their agentic AI investments.
Among banks, the top processes targeted for AI agent deployment are customer service (75%), fraud detection (64%), loan processing (61%), and customer onboarding (59%). Insurers show a similar pattern: customer service leads at 70%, followed by underwriting (68%), claims processing (65%), and onboarding (59%).
The report identifies 5% of surveyed firms as having already built the capabilities to run agentic AI at true enterprise scale. According to Capgemini, those firms share three traits: they coordinate agents across both business and technology functions, anchor deployments to a trusted cloud foundation, and connect AI spending to measurable business outcomes.
Cloud infrastructure has become a central concern for firms running multi-agent systems at scale. Fifty-two percent of respondents said they view hyperscalers as their preferred cloud partner for governing multi-agent deployments, citing requirements around regulatory compliance, data residency certifications, security frameworks, and auditability. Ninety percent said they are using or planning intelligent model routing to control costs and optimize performance as token expenditure grows.
Nearly half of financial institutions are creating new roles to oversee AI agents. Only 33% are building proprietary agents in-house, suggesting most firms still rely on vendor solutions.
Ravi Khokhar, Capgemini’s Global Head of Cloud for Financial Services, said in the report that financial institutions have moved quickly from individual AI agents to coordinated multi-agent systems spanning entire business lines, but that scale without proper infrastructure, governance, and a clear line to commercial impact delivers little value.
Capgemini estimates AI agents could generate up to $450 billion in economic value for the financial services industry by 2028. The share of executives prioritizing business outcomes in AI funding decisions is projected to rise 141% by that year.
Copy link