Visa Tests AI Agent-Initiated Payment Systems in Europe

Visa is preparing its payment systems to support transactions initiated by artificial intelligence (AI) agents, marking a significant evolution in the way banking infrastructure handles purchases. The company’s 'Agentic Ready' programme, now rolling out in Europe, is a pilot designed to test how financial institutions respond when software—not a human—initiates transactions.

Traditionally, payments rely on an individual making a decision to purchase, with banks and card networks validating identity and intent. However, advances in AI suggest that, in some cases, software agents may soon be able to independently complete transactions on behalf of their users after receiving goals or rules.

Visa is working with banks including Commerzbank and DZ Bank to determine how AI-driven transactions fit into today’s tightly regulated banking environment. According to information from Visa, the effort focuses on enabling secure transactions where AI systems act as the initiating party—meaning an AI agent, rather than a customer, could carry out purchases within user-defined boundaries.

A core challenge for banks is verifying intent and identity in this new model. Current payment systems are built around human involvement; if AI agents begin to initiate payments, systems must establish methods for authenticating software agents and defining their level of autonomy. For example, an AI might monitor inventory, compare prices, and automatically purchase supplies once predefined conditions are met.

Visa has drawn parallels between this shift and the banking sector's adaptation to online payments in past decades, when new types of transaction flows demanded revised controls and compliance measures. Participating banks, including Commerzbank and DZ Bank, are scrutinizing how AI could be embedded in current systems without compromising legal and regulatory requirements—particularly regarding fraud prevention, audit trails, and explicit customer consent.

The company emphasises that this work is focused on payment infrastructure, rather than consumer-facing tools. It involves developing methods to authenticate software agents, approve their transactions, and manage potential disputes. As these concepts move from pilot to potential wider adoption, banks will also need to address technical and procedural standards to mitigate risk.

In the broader context, large organisations already employ software to automate routine back-office tasks, and many are investing in AI strategies to reduce costs and enhance efficiency. The use of AI agents for procurement, for example, could streamline routine purchases while raising new questions about oversight and limitations.

Regulators are closely monitoring AI’s integration into financial decision-making, especially concerning credit checks and fraud management. A recent report by RepRisk notes that AI-linked incidents are becoming more frequent and costly for banks, suggesting the need for enhanced controls as AI adoption grows.

Visa’s current phase is largely experimental, focused on system design and testing. As AI agents take on more transactional responsibilities, banks and payment networks will be challenged to adapt to users who are not people, but autonomous pieces of software.

[source: artificialintelligence-news.com]