The rise of agentic payments: When treasury systems become autonomous bankers

AI-powered treasury agents are reshaping payments, challenging banks to build a new digital trust architecture.

The treasurer has left the chat

Corporate treasury is entering a new phase of digital evolution. While automation, APIs, and straight-through processing have become standard, the next wave of innovation is being driven by agentic AI embedded directly within treasury management systems (TMS). Leading providers are beginning to introduce agentic AI-powered capabilities that go beyond reporting and insights. These capabilities enable treasury platforms to analyze liquidity, recommend actions, and orchestrate workflows with minimal human intervention.

As agentic AI-driven treasury platforms become the primary decision layer for organizations, banks face an uncomfortable reality: user experience alone will no longer differentiate them. When agents rather than humans choose rails, products, and liquidity options, traditional banking payments expertise might lose some of its value to AI agents. This expertise includes knowledge about rail characteristics, cutoffs, corridor pricing, correspondent chains, format quirks, value-date arithmetic, and operations.

Rather than relying solely on traditional channels such as ERP integrations, host-to-host connectivity, APIs, or batch file transfers, banks will need to support a new class of interactions. These interactions will be initiated by treasury agents acting on behalf of corporate users.

Such AI agents may request liquidity information, evaluate funding options, initiate treasury actions, or trigger payment workflows based on predefined policies and governance controls.

This shift creates a fundamental challenge for transaction banks: how to securely enable machine-to-machine financial interactions while preserving trust, accountability, and regulatory compliance.

Know your agent: The new digital trust architecture

Know your agent (KYA) applies know your customer (KYC) principles to AI agents, requiring banks to verify an agent’s identity, authority, and actions just as they would for a human customer. The key challenge goes beyond granting access. Banks need clear, auditable proof that the agent was authorized to act on behalf of a client for a specific purpose and session.

Retail payments have already begun addressing this reality through agentic protocols. Examples include Google’s AP2, Visa’s Trusted Agent Protocol, and Mastercard’s Agent Pay. These protocols all assume a natural-person principal, a card rail, a merchant counterparty, and a network rulebook with an adjudication mechanism behind it. Corporate treasury has none of these four elements.

Corporate banking is still in the early stages of understanding what machine-initiated financial activity means. Despite significantly larger transaction values and more complex approval structures, it lacks an equivalent banking framework or protocol.

Corporate banking faces a significant challenge because many authorization and control processes were designed for human users. It often depends on offline authorization, signature verification, and approvals in addition to credentials and access controls. As a result, banks need to establish how AI agents can obtain and demonstrate the same level of trusted, auditable authority to act on behalf of corporate clients.

Banks must now answer a new question: How do we verify who the customer is, which AI agent is acting on their behalf, and the authority under which that agent is acting?

Engineering the trust layer for agentic banking

We see this as the next frontier of trust in financial services. Just as KYC became a foundational banking capability, KYA will become critical for enabling safe, scalable agentic banking.

The real requirement goes beyond agent identification. Banks and corporations need machine-readable mandates that clearly define an agent’s authority. These mandates need to specify what an agent can do and which accounts and payment corridors it can access. They also need to define transaction limits, authorization hierarchies, and real-time revocation capabilities.

Corporate banking therefore needs a new digital trust architecture that goes beyond authentication alone.

The time to build is now

Banking leaders have a narrow window to act before corporate AI policies become embedded by treasury technology vendors and enterprise software providers. Waiting for widespread demand may result in banks losing influence over how corporate treasury agent ecosystems are designed.

For banking executives, the key questions are no longer theoretical:

• Can a corporate treasury agent discover your services and choose them based on the data and capabilities you expose digitally?
• Can your bank verify and control machine-initiated instructions today?
• How quickly can you revoke an AI agent’s authority across channels?
• What liability model exists when an agent acts outside its mandate?

The transition to agentic treasury is already underway. Banks that treat AI agents as just another channel risk commoditization. Banks that build trusted agent ecosystems can create entirely new sources of value. AI agents will become part of treasury operations. The question is whether your bank will be ready when they do.

Related research and insights

Expert perspectives