Ready for agentic commerce.

Agentic commerce is a form of digital commerce in which an AI agent can support or carry out parts of a buying journey on behalf of a consumer or business. An agent may help search, compare and select an option. With the right instructions and approval, it may also initiate or support the payment.

An agent-driven payment is a payment journey in which an AI agent participates in initiating or supporting a transaction, based on instructions or conditions set by a consumer or business.

Yes, as live, production proof points Worldline has completed end-to-end agent-driven payments with ING and Mastercard, with ING and Visa, and with Crédit Agricole and Mastercard, across multiple European markets. These show the payment layer can operate through established infrastructure today. Broader commercial availability is still developing, and will depend on continued alignment between merchants, banks, payment networks, technology providers and regulators.

Never. The models Worldline has demonstrated are designed around consumer instructions and approval. The agent supports or acts within defined conditions, while the bank retains authentication and authorisation. We're also exploring how agents could make more autonomous purchases without direct involvement of the customer, based on high-level intent.

SCA stays relevant when an AI agent participates in a payment journey. The consumer still needs an appropriate way to authenticate and approve the transaction, and the issuing bank retains control over authorisation. In Worldline and ING's work with Visa, the consumer confirmed purchase intent biometrically using Visa Payment Passkey, supporting a journey designed around established authentication requirements.

Biometric authentication gives the consumer a familiar, low-friction way to confirm a purchase within an agent-driven journey. In Worldline and ING's work with Visa, Visa Payment Passkey was used to authenticate the consumer's purchase intent.

A merchant needs a clear, verifiable record of what the customer approved — the product or service, merchant, amount, timing and any other conditions. The payment journey should preserve this approval context so the bank and other participants can understand what was authorised, by whom and under which conditions.

An agent-driven transaction may need additional context alongside standard payment data: how the agent participated, what the customer approved, the applicable purchase conditions, the authentication event and the final outcome. This supports issuer decision-making, traceability and dispute handling.

AI agents shouldn't need access to raw payment credentials. Tokenisation can replace sensitive card information with payment tokens used within controlled journeys. Authentication, token controls and secure processing keep the agent separate from the underlying card credentials.

Agent-driven payments should operate within the security, authentication, fraud management and regulatory controls used for digital payments today. New risks can arise from unknown agents, altered instructions or unusual transaction behaviour. The consumer's approval, agent and transaction signals, issuer controls and a clear record of the journey all help manage them.

Agent-driven payments continue to use established payment rails and their dispute processes. The key difference is the potential need for additional evidence showing the agent's role and the conditions the customer approved. The exact allocation of liability depends on the payment method, scheme rules, contractual arrangements and applicable law.

Yes, some models already support delayed completion within clearly defined conditions. In the Visa transaction Worldline and ING demonstrated, the model allowed completion up to 48 hours after authorisation. In future, agents may complete transactions with greater autonomy based on a customer's high-level intent.

Merchants may need to make products and buying journeys understandable to AI agents, while keeping checkout, approval and payment confirmation reliable and secure.

Banks may need to process new transaction information, authentication patterns and risk signals, while retaining control over authorisation and consumer protection.

Yes. Worldline began researching agent-initiated payments in 2021 and continues to publish on the topic. Worldline experts co-authored AI Agents in Payments: Applications, Risks and Regulations with HEC Paris, published in the European Journal of Risk Regulation by Cambridge University Press.

Worldline is currently working with ecosystem partners to test and develop the payment capabilities agent-driven commerce needs. The live transactions are proof points and shouldn't be read as broad commercial availability.

Yes. The Worldline MCP Server connects our payment functionality to AI agents through the Model Context Protocol, so developers can build and test AI-powered payment workflows using natural-language prompts. It's available now via our developer portal, ConnectAI, as a first building block toward agent-initiated payments.