Visa prepares payment infrastructure for transactions initiated by AI agents

Visa is preparing its payment infrastructure for transactions initiated by AI agents. The company is already testing in Europe how such transactions can be conducted securely without disrupting traditional banking processes.

What Visa is testing

On March 17, 2026, the company launched the Visa Agentic Ready program in Europe. This is not a separate consumer service, but a working framework for banks and payment partners who need to understand how to process transactions initiated by software rather than by a person in an app or on a website. The first wave includes more than twenty issuers and partners, including Commerzbank, DZ Bank, Barclays, HSBC UK, Revolut, and Banco Santander. Tests are being conducted jointly with Visa and selected merchants in a controlled environment that increasingly resembles real-world conditions.

In essence, Visa is preparing for a model in which AI does not simply suggest what to buy, but completes the purchase. A user or company sets a goal, constraints, and rules, while the agent itself searches for the right product, compares options, and initiates payment. For card networks, this is a significant shift: previously, payments almost always began with explicit human action, but now software can be the initiator. This changes the very point where a bank must verify consent and authority limits.

"The stronger AI agents influence how people choose and buy goods, the faster payments must also change,"

Visa Europe stated.

How this should work

Today, a card transaction relies on simple logic: there is a specific cardholder, their intention, and purchase confirmation. In agentic commerce, intention is still set by the person, but part of the operational work goes to the agent. For example, it can monitor inventory levels, find the best price within a specified limit, and place an order without requiring a manual click each time. Banco Santander has already demonstrated a practical scenario where a book purchase was completed in a real-world environment using a Visa card.

Within Agentic Ready, banks are testing not just one function, but the entire path of such an operation:

• card binding and token issuance for agent actions
• authentication through modern mechanisms like passkeys and biometrics
• payment authorization taking into account limits, rules, and purchase type
• consent recording, action logs, and dispute control
• transaction behavior with real merchants, not just in a test sandbox

Visa describes this as a trust layer—a layer of trust where tokens, identity, risk models, and user constraints converge. The idea is for the agent not to receive a "free card," but to operate within a narrow corridor of authority. If a person has authorized searching for airline tickets up to a certain amount or regularly purchasing office supplies, the system should be able to verify exactly this context, not just the existence of an active card.

Where the bottlenecks are

The main question here is not whether AI can click a "pay" button. The main question is how to prove to a bank and payment network that the agent truly acts on behalf of a specific customer, within their consent, and without exceeding their authority. This drives interest in tokenization, biometric authentication, action logs, and additional rules on the issuer's side. Commerzbank and DZ Bank are precisely testing how to embed such scenarios into existing processes without conflict with compliance requirements.

There is also a practical dimension. If an agent makes a mistake with a product, buys at an unexpected price, or exceeds an allowed category, the bank will still have to investigate the dispute, reconstruct the chain of actions, and resolve the liability question. Therefore, the current stage is not a mass rollout of "autonomous shopping for everyone," but the development of infrastructure, audit, and control. For corporate purchases, this approach is particularly interesting: it can eliminate some routine work, but simultaneously requires very clear limits and rules.

What this means

The payment market is preparing for the moment when AI becomes not a consultant, but a full participant in a transaction. If Visa's pilots are successful, banks will begin issuing agents limited and verifiable payment authority for routine purchases. And this is no longer a distant concept: on April 29, 2026, Visa announced the expansion of Agentic Ready beyond Europe, demonstrating how quickly the topic is transitioning from experiment to industry standard.