The future of finance is autonomous. Agentic commerce, where AI systems execute complex transactions and financial decisions with minimal human intervention, is the next frontier. For large financial institutions (FIs) burdened by decades of legacy payment infrastructure, bridging this gap is not just an aspiration—it’s an imperative for survival and competitive relevance. Success hinges on a phased, strategic approach focused on data, APIs, and governance.
1. Modernizing the Data Foundation
The biggest blocker to AI agents isn’t the AI itself, but the data readiness. Legacy systems often silo data, making it inconsistent, fragmented, and difficult for AI agents to access in real-time.
- Real-Time Data Pipelines: FIs must invest in modernizing data pipelines to aggregate data from disparate core systems, transforming it into a unified, real-time data fabric. This is the essential fuel for an autonomous agent.
- Quality and Context: Data quality controls and robust lineage—tracking data from origin to use—are non-negotiable. An agent’s decision is only as good as the information it receives, and in a heavily regulated industry, explainability and auditability start with the data.
2. The API-First Strategy: Creating an Integration Layer
A full “rip and replace” of core payment systems is too risky and costly. The practical solution is an API-first intermediate layer.
- API Gateway: Develop a comprehensive set of Application Programming Interfaces (APIs) that act as secure, standardized conduits to the legacy payment rails (e.g., ACH, Fedwire, SWIFT). This allows the AI agent layer to interact with the underlying infrastructure without direct integration, offering both insulation and interoperability.
- Protocols for Trust: FIs should adopt emerging open standards like the Agent Payments Protocol (AP2). AP2, or similar protocols, provides a secure, payment-agnostic framework for AI agents to transact using cryptographically signed mandates. This establishes an unrepudiated audit trail, essential for regulatory compliance and consumer trust in autonomous commerce.
3. Phased Deployment and Strategic Use Cases
The transition must be iterative, starting with low-risk, high-value operations.
- Smart Overlays: Start by “wrapping” AI agents around existing, well-defined processes that rely on legacy data and rules. For instance, deploy agents for Payment Processing Monitoring to flag exceptions and anomalies, rather than making autonomous corrections initially. This leverages the agent’s analytical power while maintaining human oversight.
- Operational Automation: Move next to complex, multi-step workflows like Reconciliation Automation or Intraday Liquidity Management. As demonstrated by early adopters, AI agents can autonomously manage cash flow decisions, optimizing the liquidity-delay trade-off in real-time gross settlement (RTGS) systems, a crucial function often tied to core legacy platforms.
4. Governance, Compliance, and Human Oversight
As autonomy increases, so does the governance burden. FIs must proactively embed control into the agent’s DNA.
- Governance-as-Code: Compliance requirements, risk appetite, and fraud prevention rules must be coded directly into the agent’s operational logic, not bolted on afterward. Every decision must be logged, reviewable, and explainable.
- Human-in-the-Loop: A robust framework for human intervention, including human override capabilities for critical decisions, ensures accountability. The operating model shifts from humans executing tasks to humans guiding, overseeing, and validating the autonomous agents.
By strategically modernizing data flows, creating an API-driven interface layer, and embedding compliance from the ground up, a large financial institution can transition from a rigid legacy payments processor to a flexible, intelligent bank capable of participating in the dynamic world of AI agentic commerce.