CaixaBank has completed a real-world payment initiated by an AI agent acting on behalf of a human cardholder, using Visa’s framework for agent-driven commerce and the existing card infrastructure that already secures ordinary transactions.
The Spanish bank said the pilot ran on Visa’s Intelligent Commerce framework and its Agentic Ready programme, using real card data on standard merchant systems. The transaction was secured by the same tokenization, identity verification, and fraud monitoring that Visa applies to conventional card payments, which the bank said together proved the viability of payments initiated by AI agents within the current infrastructure. The identity and authorization layer is central to the model, since a payment made by software on a person’s behalf still has to be tied back to a verified human who authorized it.
Visa launched its Agentic Ready programme in Europe in March to help banks test and validate agent-initiated transactions, and has run similar trials with DBS and Santander. The space is competitive: rival network Mastercard says it completed the first agentic payment in France the previous week, working with Crédit Agricole and Worldline.
CaixaBank, headquartered in Valencia and serving around 12 million cardholders, is investing 5 billion euros through a 2025-to-2027 technology strategy focused on integrating generative AI into its workflows and transaction processing. The bank set up an internal AI office late last year and has begun rolling out conversational AI agents across its digital channels, including for customer onboarding.
The pilot pushes agent-driven payments closer to the identity questions that networks are racing to answer, after Visa reported early progress on agent-initiated transactions and a trusted-agent protocol, and as Mastercard and Google launched a Verifiable Intent framework to tie a verified identity to an agent’s instructions.
Sources: FinTech Futures
–
By the ID Tech Editorial Team
