Deutsche Bank and Ant Digital Technologies Release White Paper: Agent Payments in Wholesale Finance
on November 9, Deutsche Bank and Ant Digital Technologies jointly released the whitepaper "Agentic Payments in Wholesale Finance," discussing the practical applications of AI agents in institutional finance and the governance framework required for responsible deployment.
The whitepaper outlines three key scenarios: treasury and liquidity management, where agents can autonomously predict liquidity gaps and surpluses, evaluate response strategies, and execute internal transfers, investments, external financing, and foreign exchange operations; cross-border payment orchestration, which weighs exchange rates, payment channels, and execution timing to optimize cost, speed, availability, compliance, and settlement risk; and trade finance, where agents can interpret complex trade documents, identify data discrepancies, coordinate exception handling, and generate financing or payment instructions once conditions are met and approvals are obtained.
On the governance front, the two institutions advocate for "bounded agent execution" rather than full autonomy, adopting a three-tier reference architecture proposed by the IMF: Intent & Orchestration, Control & Authorization, and Settlement, which separates AI reasoning from transaction execution. Supporting requirements include explicit, limited, and revocable authorization limits, binding agent identities to accountable entities, and "Know Your Agent" (KYA) verification. The whitepaper also cites Ant Digital Technologies' agent platform, Anvita Flow, as a real-world implementation case.
In the foreword, Sabih Behzad, Head of Digital Assets and Currency Transformation at Deutsche Bank, stated that the value lies not just in automated payment execution, but in improving how decisions are made, coordinated, and controlled. The president of Ant Digital Technologies emphasized in the same foreword that agents require verifiable identities, explicit and revocable authorizations, and a clear separation of decision-making, control, and execution.