Visa has begun a pilot program in Europe that allows artificial‑intelligence agents to initiate payment transactions on behalf of users. The initiative, called “Agentic Ready,” is being carried out in partnership with German banks Commerzbank and DZ Bank. The goal is to adapt existing card‑network infrastructure so that software agents can search for products, evaluate conditions, and complete purchases without direct human confirmation.
How the Pilot Works
In the current payment model, a transaction is triggered by a person who authorises a purchase. Visa’s pilot introduces a new layer in which an AI system, once given a goal or set of rules, can act as the initiating party. The system would need to prove that it is acting on behalf of a user and that it has the appropriate level of autonomy. The pilot focuses on secure transactions, ensuring that identity verification and intent are maintained at the system level.
The program envisions routine or repeat purchases being handled automatically. For example, an AI agent could monitor inventory levels, compare prices across suppliers, and execute a purchase when predefined thresholds are met. This approach mirrors the early shift to online payments, which required banks to adjust to a new transaction flow.
Compliance and Risk Management
Commerzbank and DZ Bank are testing how AI agents can be integrated without violating regulatory requirements. Key compliance areas include fraud detection, audit trails, and customer consent. The banks must demonstrate that any automated transaction still meets oversight standards. A RepRisk report cited in the pilot’s documentation notes that banks are already experiencing more frequent and costly AI‑related incidents, with potential losses reaching multi‑million‑dollar levels.
Visa’s work is focused on infrastructure rather than consumer‑facing tools. The company is defining how agents are authenticated, how transactions are approved, and how disputes are handled if an error occurs. These measures aim to preserve the integrity of the payment network while allowing new forms of transaction initiation.
Implications for Enterprise Procurement
Large organisations often require multiple approval steps for purchases. AI agents could streamline this process by handling routine transactions within set limits. While this could reduce manual effort, it also raises the need for clear rules governing agent behaviour. Without such controls, the risk of errors or misuse increases. Regulators are paying closer attention to AI decision‑making in areas such as credit assessment and fraud detection, which may influence how quickly enterprises adopt automated procurement.
Future Outlook
The current phase of the program is limited to testing and system design. If the pilot proves successful, payment networks may need to accommodate a new type of user that does not hold a physical card but can still initiate purchases. Banks will likely continue to set strict rules, monitor activity, and manage exceptions. The broader industry will watch closely to see whether AI‑initiated transactions become a standard feature of payment systems, potentially reducing the need for direct human input in everyday transactions.
