Worldline, ING and Visa show a live “agentic” card payment in Europe, using biometric and customer authentication

Worldline, ING and Visa said they have completed a live, end-to-end “agentic” payment transaction in Europe, highlighting how AI-assisted payment initiation could work in production systems without skipping key security steps. The demonstration was carried out in Germany and was presented as a proof point for “agentic payments,” a term used for payment flows where software agents can take part in tasks like initiating a purchase after interpreting customer intent, rather than relying only on manual, step-by-step user input.

In the described example, the companies emphasized that the transaction still relied on strong customer verification, including biometric authentication alongside customer authentication. The reporting framed the result as a way to show that agentic payment concepts can be implemented within existing authentication expectations, rather than treated as a purely experimental shortcut.

The use of biometric factors matters because payment fraud and account takeover remain central concerns for card networks and issuing banks. In practice, biometric authentication is one of several mechanisms used to confirm a customer’s identity at the moment of authorization. By tying the “agentic” part of the flow to biometric and other customer checks, the demonstration aimed to address a common hesitation around whether AI-driven initiation could broaden risk.

According to the coverage, the demonstration was intended to show that agentic payments can operate securely in a live setting, rather than only in a lab. The companies did not, in the publicly referenced materials available here, disclose architecture-level details such as which AI agent orchestration layer was used, the specific biometric method, or how the payment authorization rules were encoded beyond the emphasis on “strong customer and biometric authentication.”

Broader context: Visa operates in a highly regulated payments environment where authentication and authorization controls are central to risk management. Issuers like ING must balance customer experience with regulatory and fraud controls, which is one reason card networks have pushed interoperability and authentication standards over the past several years. An “agentic” workflow, if deployed more widely, would need to fit into these constraints and still produce the auditability and control that payment systems require.

For Worldline, the demonstration also speaks to its role as a payments and merchant-facing technology provider. If AI-assisted initiation becomes more common, payment service providers may be expected to support the integration patterns needed for those flows, including the secure handoff between an AI agent’s intent and the underlying payment authorization steps.

The companies did not provide, in the referenced posts and summaries available for this report, a full description of transaction limits, fraud scoring outcomes, acceptance rates, settlement performance, or compliance documentation. That means it is not yet clear how the transaction performed under stress conditions, whether the example represents a repeatable rollout across corridors, or what additional onboarding would be required to scale from a demonstration to a broader commercial deployment.

Going forward, market participants will likely watch for follow-on announcements that specify what is being piloted or rolled out, which merchants or partners are involved, whether tokenization and authorization flows change, and how regulators view agentic initiation when it is paired with authentication indicates like biometrics. The companies’ next public disclosures may clarify whether this is a one-off interoperability test or the start of a broader product trajectory for AI-enabled payments.