Multi-Agent AI Enters Banking Core Operations But 90% Struggle With Enterprise Orchestration

Financial Services Firms Deploy Multi-Agent AI Despite Orchestration Gaps

Financial institutions are moving multi-agent AI from experimental phases into core banking and insurance workflows, according to Capgemini's World Cloud Report for Financial Services 20271

1

2

2

. The research, based on a survey of 1,100 financial services executives across 14 markets, shows that 59% of firms are piloting or deploying multi-agent AI systems in production2

2

. Yet only one in ten have achieved the full orchestration required to run them autonomously across the enterprise, exposing a significant gap between adoption and operational maturity.

ROI Challenges Drive Shift Toward Business Outcomes Over Efficiency

Nearly half of financial services firms cannot consistently demonstrate a return on their agentic AI initiatives, while 51% lack a clear framework for assessing whether their AI investments deliver sufficient value to justify continued spending1

1

. This ROI uncertainty is forcing a strategic pivot. By 2028, the number of executives prioritizing measurable business outcomes as their primary AI funding criterion is projected to increase 141%, while those focused on efficiency will drop 42% over the same period1

1

2

2

. Financial institutions are implementing disciplined cost and performance controls to balance the capabilities of multi-agent AI systems with deployment expenses. To address these challenges, 90% are using or planning to implement intelligent model routing to direct tasks toward models offering the most appropriate balance of accuracy and cost1

1

2

2

.

Cloud Governance Emerges as Critical Foundation for Multi-Agent AI

Cloud infrastructure is becoming the operational foundation for agentic AI at scale, with 52% of banks and insurers identifying hyperscalers as their preferred partners for governing multi-agent deployments1

1

2

2

. Financial institutions increasingly recognize cloud as the common execution layer that governs how agents access data, make decisions, and take action. Demand is consolidating around major providers driven by requirements for regulatory compliance certifications, security and identity frameworks, and auditability at scale2

2

. Eight in ten executives identify reducing technology debt as a priority for scaling agentic AI at scale1

1

2

2

, recognizing that modernizing their underlying tech stack is a prerequisite to succeed in the agentic AI era.

Customer-Facing AI Agents to Transform Core Workflows Within Three Years

AI deployment in financial services is expanding into customer-facing responsibilities, with three quarters of respondents expecting agents to become a key part of customer engagement within three years1

1

2

2

. Account servicing, loan origination and underwriting, and product recommendations are among the core workflows expected to see greater adoption. This shift toward autonomous operations while meeting regulatory requirements reflects the increasing decision-making responsibilities being assigned to AI agents in banking and insurance. Hybrid sourcing, combining in-house development, packaged solutions, and external partners, is expected to become the dominant approach for more than two-thirds of financial institutions within three years, up from 36% today1

1

2

2

.

Elite 5% Achieve Enterprise-Wide Orchestration Across Operations

Capgemini identifies a small group representing just 5% of respondents that have already established the capabilities required to operationalize agentic AI at enterprise scale1

1

2

2

. More than half of these leaders orchestrate multi-agent systems across operations, technology, and customer experience, compared with just 8% of other firms1

1

. Additionally, 83% link AI investment directly to business results. Ravi Khokhar, global head of cloud for financial services at Capgemini, said: "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"1

1

2

2

. He emphasized that scaling these systems requires end-to-end orchestration across operations, governance, and security, alongside a clear link to bottom-line impact. The few pulling ahead are treating cloud foundation not as a commodity, but as the foundational pillar of their agentic AI strategy.