Financial firms race to deploy multi-agent AI systems

Capgemini reports that financial institutions are moving from isolated AI agent trials to multi-agent AI systems across core banking and insurance workflows. Its latest research found that 59% of firms are piloting or deploying such systems in production.

The findings suggest a shift in how banks and insurers approach AI, with firms increasingly tying investment decisions to business results rather than operational efficiency alone. At the same time, only 10% of respondents said they had achieved enterprise-wide orchestration for autonomous multi-agent systems.

The study was based on a global survey of 1,100 C-suite executives at financial services firms across 14 markets, covering retail banking, wealth management, payments, corporate and investment banking, and insurance.

The industry is under pressure to show a clearer return on AI spending. About half of the financial services firms surveyed said they cannot consistently demonstrate returns from agentic AI initiatives and do not yet have clear frameworks to assess whether those projects justify continued investment.

That concern is changing funding priorities. By 2028, the number of executives who expect outcomes to be their main criterion for AI funding is projected to rise by 141%, while the number focused primarily on efficiency is projected to fall by 42%.

Cloud focus

Cloud is becoming the main execution layer for multi-agent systems, as institutions seek environments that can manage how AI agents access data, make decisions and take action. More than half of the banks and insurers surveyed said hyperscalers were their preferred cloud partners for governing multi-agent deployments.

Security, regulatory and audit requirements are central to that choice. The research found that 52% of executives prefer hyperscalers as cloud partners for governing agentic AI, citing compliance, security, auditability and data residency among the main considerations.

Financial institutions are also paying closer attention to cost controls as AI systems grow more complex. The report found that 90% of firms are using or planning to implement intelligent model routing to direct tasks to the most cost-effective and accurate models, partly to manage token costs.

The findings also highlighted the need for broader technology modernisation. Eight in ten executives said reducing technical debt was a priority, reflecting concern that older systems may limit the ability to scale AI tools safely and consistently.

Business uses

The research suggested that AI agents are moving into more business-critical roles across customer-facing and operational functions. Half of executives expect AI agents to become part of the enterprise workforce by 2029, while 76% expect them to play a key role in customer engagement within the next three years.

Respondents pointed to several early use cases, including account servicing, cited by 51% of executives, loan origination and underwriting at 50%, and product recommendations at 47%.

These figures suggest a broader role for AI in day-to-day financial services operations, but the report also indicated that few institutions have reached full organisational maturity. Capgemini identified a small group representing 5% of firms surveyed that had institutionalised the structures needed to run agentic AI at enterprise scale.

According to the research, those firms stand out in three areas: coordinating agents across business and technology functions, building a cloud foundation for autonomous execution, and linking AI spending to measurable outcomes.

Hybrid model

Firms are also increasingly adopting a hybrid sourcing model for agentic AI. That approach combines in-house development for differentiation, packaged solutions for faster deployment, and external partners for access to wider ecosystems.

Capgemini expects this to become the dominant approach for more than two-thirds of financial institutions within three years. The finding reflects a market in which firms are trying to balance control over strategic systems with the speed and scale offered by outside providers.

Ravi Khokhar, Global Head of Cloud for Financial Services at Capgemini, said the industry had moved quickly in a short period.

"In just 12 months, financial institutions have moved beyond experimenting with individual AI agents tackling isolated tasks to deploying multi-agent systems that coordinate actions across entire business lines. Yet adoption alone does not create value," said Ravi Khokhar, Global Head of Cloud for Financial Services at Capgemini.

"Successfully scaling these systems requires infrastructure that drives end-to-end orchestration, effective governance and security by design, and a clear link to bottom-line impact. Cloud is now the operational foundation that makes all this possible. The few pulling ahead are not treating cloud as a commodity, but as the foundational pillar of their agentic AI strategy," Khokhar said.