At the Visa Payments Forum (VPF) in Paris, Visa announced that agentic commerce is moving out of closed test environments: the company says that AI agents have already executed live transactions across Europe. This year’s VPF was held in early July in Paris; about two thousand industry delegates attended and a demo day gave press and partners a chance to see the company’s strategic plans and products.
Visa’s core announcement is that AI agents participating in the Agentic Ready program are performing real purchases at independent merchants on behalf of cardholders. These transactions proceed much like human shopping: the agent browses offers, asks follow-up questions if needed, and completes purchases within user-defined budgets and rules.
Visa stresses that consumer explicit consent is required for every agent-initiated purchase, and that agents operate under parameters set by the user. The technical and operational framework behind these capabilities is provided by Visa Intelligent Commerce, the company’s portfolio of initiatives designed to support scalable, secure AI-driven commerce.
How it works in practice
In Visa’s demo the agent experience was illustrated via a hypothetical toggle in the ChatGPT app — the June 10 OpenAI–Visa cooperation was noted during the forum — where a user activates a personal Visa agent, assigns a task (for example, planning a long weekend in Milan), and configures price and preference sliders. The agent returns flight, hotel and activity suggestions and, once the user selects an option, completes the online purchase with payment approval.
Agents can use prior transaction histories and loyalty data to personalise decisions; Visa says the more relevant data available, the better an agent can represent the user’s tastes and routines.
Who controls the transactions and how are agents authenticated?
Visa emphasises that trust and control are central: every AI-initiated transaction must be authenticated. The company developed Visa Payment Passkeys to ensure that agent-initiated transactions are tied to verified cardholders’ approvals. Visa says that in the live transactions executed so far both the AI agents and the cardholders were authenticated via Passkeys, aligned with Europe’s strong customer authentication requirements.
The Visa Trusted Agent Protocol (TAP) and the Agent Directory are intended to let merchants automatically and reliably identify which AI agents are authenticated. TAP is designed to integrate with existing risk-management and regulatory systems and is supported by infrastructure partners such as Cloudflare and Akamai to ease merchant integration.
Visa executives, including Oliver Jenkyn, Group President, acknowledged psychological barriers: handing over financial, medical and card data to an AI agent is a behavioural hurdle. Their research points to lower initial trust among women compared with men when it comes to delegating purchasing control to AI.
Liability, risks and cybersecurity
Visa also introduced the Visa Threat Intelligence Platform (VTIP) to help banks and fintechs detect and respond to cyberattacks and fraud earlier. Visa reports that it blocks some 90 million cyberattacks and 11 million phishing emails each month across more than 200 countries, and that it has invested over 13 billion dollars in technologies over the past five years to reduce fraud and strengthen its network.
VTIP aims to surface financial-sector-specific threat indicators, uncover security gaps, detect impersonation and brand abuse, and transform compromised payment data found on the dark web into actionable intelligence for banks’ risk teams.
Which banks and when will consumers see this?
Visa says more than thirty European issuers are already participating in the Agentic Ready program; the list includes Abanca, Alpha Bank, Bankinter, Barclays, BBVA, CaixaBank, Ceska Sporitelna, comdirect, Commerzbank, Cornèrcard, Bank of Cyprus, Deutsche Kreditbank, HSBC UK, ING, Klarna, Lloyds Banking Group, mBank, Nationwide Building Society, NatWest, Nexi Group, Nordea Bank, OP Cooperative, Piraeus Bank, PKO Bank Polski, Revolut, S-Payment, Swisscard, Tatra banka, Banca Transilvania, Vanquis Bank and Viseca. No Hungarian banks were listed among participants at the time of the announcement.
On timing, Philip Konopik, Senior Vice President at Visa Europe, said the rollout remains uncertain: after previously predicting a 12-month horizon, he now characterises broader adoption as likely by the end of the year. Visa highlights that full-scale adoption requires coordination across the ecosystem — issuers, merchants, infrastructure providers and standards bodies — so that trust is embedded from the outset.
Visa also notes that merchants must adapt: they need processes to identify authenticated AI agents and to distinguish trusted AI-driven interactions from unauthenticated traffic.
Private concierge for wealthy clients
Visa showcased an AI-powered concierge proposition targeted at banks and private-banking clients, particularly high-net-worth individuals (HNWI) and ultra-high-net-worth individuals (UHNWI). Within the Visa Infinite programme, cardholders can receive AI-concierge services that arrange personalised and exclusive experiences on their behalf. Visa says its definition of affluent customers covers the top 10 percent by wealth, who account for roughly 45 percent of consumer spending and spend at least 25,000 dollars per year at the lower end of that band.
Visa named several partners in this area and said that Barclays, Privat3 Money, Isbank and HY10 have joined the renewed Visa Infinite partner roster so far.
Why is Visa investing in agentic commerce?
Visa frames this work as part of its transformation from a traditional card network to a leading digital payments platform. As payments become more digital and AI takes a larger role in decision-making, Visa views agentic commerce as a strategic opportunity to embed its platform at the centre of future transactions.
The company underlined its broader investment posture: it has seen about a 50 percent increase in its stock price over five years and points to multi‑billion-dollar investments in AI and security. Visa’s moves are in line with competitors’ efforts — for example, Mastercard and Santander have also executed AI-agent payment tests under Mastercard’s Agent Pay initiative.
Conclusion
At the Paris VPF Visa presented agentic commerce as having entered live, real-world testing: AI agents in the Agentic Ready program have already completed transactions across Europe. The company introduced authentication and identification tools (Visa Payment Passkeys, Trusted Agent Protocol, Agent Directory) and a cyber threat intelligence platform (VTIP) to protect the ecosystem. Visa insists that consumer approval and security remain prerequisites for agent-initiated payments, while noting that significant technical, regulatory and behavioural challenges must still be addressed before widespread adoption.
