Deutsche Bank and Ant Digital Technologies jointly released a white paper today: “Agentic Payments in Wholesale Finance.”

This isn’t just empty talk about “AI getting things done.” The white paper identifies three use cases: liquidity management (anticipating shortfalls and making transfers, investments, financing arrangements, or FX transactions within authorized limits); cross-border payment orchestration (weighing exchange rates, payment channels, and timing); and trade finance (reading documents, spotting discrepancies, and issuing instructions only after approval).

The key principle is “bounded agent execution”: intent orchestration, authorization controls, and settlement are separated into three layers. AI can propose a plan, but it cannot grant itself broader authority. Authorization must be explicit and revocable, the agent’s identity must be tied to an accountable entity, and organizations must also practice “Know Your Agent” (KYA). One implementation is Ant Digital Technologies’ Anvita Flow.

Sabih Behzad, Deutsche Bank’s head of digital assets, put it plainly: the value isn’t in automating payments, but in how decisions are made, coordinated, and controlled. In institutional payments, agents that can take action have arrived—and respecting boundaries is the key.
#AIAgent #AgenticPayments