There has been a missing piece in the agentic commerce story for the past two years. AI agents could browse. They could compare. They could find the best deal across a hundred merchants in seconds. But completing a purchase securely, with proper identity verification, spending controls, and fraud protection built in, required infrastructure that no major payment network had formally opened to agents.
On April 8, 2026, Visa changed that.
With the launch of Intelligent Commerce Connect, Visa became the first major card network to provide a formal, protocol-agnostic entry point for agent-initiated transactions, with tokenization, spend controls, and fraud authentication built directly into the payment layer.
This is not a roadmap announcement. It is live infrastructure, in pilot with real partners, and it changes the calculus for every merchant operating at scale.
The fundamental problem with AI agent payments has always been trust, not technology. Giving an AI agent the ability to spend money meant giving it access to payment credentials. That model was never going to reach mainstream adoption.
Visa's approach solves this elegantly. When an AI agent initiates a purchase, Intelligent Commerce Connect identifies the card being used, replaces the real card number with a secure token, and routes the transaction to the appropriate payment network. Before passing the credentials to the merchant, it verifies that the agent is operating within the consumer's predefined spending limits. The transaction is logged for full transparency.
The card never touches the agent. If the agent is compromised, the token is revoked. The underlying account is untouched. This is the trust architecture that agentic commerce needed, and Visa built it on top of its existing fraud detection infrastructure, which already processes billions of transactions annually.
The platform supports the four major agent protocols active in the market today:
That decision to be protocol-agnostic is strategic.
Rather than betting on a single standard winning the ongoing protocol wars between Google, OpenAI, and others, Visa built a layer that works across all of them. For merchants, that means a single integration point regardless of which AI platforms their customers end up using.
By making one hub that connects businesses to multiple agentic protocols and payment rails, Visa is effectively doing for agentic commerce what it once did for e-commerce acceptance more broadly: providing common rails that allow innovation to scale.
Most of the coverage around Intelligent Commerce Connect has focused on the payment side. The more strategically important feature for merchants is the catalog layer.
The platform helps merchants make their product inventories and product details, including descriptions, specifications, and prices, accessible so consumers can discover, select, and check out within the AI platform experience.
This is the distribution question that merchants need to be asking right now. If an AI agent is shopping on behalf of a consumer and your product catalog is not accessible to that agent, you are not in consideration. The merchants who get their catalogs discoverable to AI platforms early will have a structural advantage in this channel before the volume becomes significant enough to fight over.
Visa's own research shows that nearly half of US shoppers, 47%, already use AI tools for at least one shopping task, from price comparisons to personalized recommendations. Visa predicts that millions of consumers will use AI agents to complete purchases by the 2026 holiday season.
That is not a five-year projection. That is this year.
Visa's launch is a forcing function. It makes visible a set of infrastructure gaps that have existed in most enterprise payment stacks for years but have not mattered until now.
Agent-initiated transactions arrive without the behavioral signals that current fraud and authorization systems rely on. There is no browsing session. No device fingerprint built up over time. No biometric data. A risk model built exclusively around human transaction patterns will produce more false declines on legitimate agent transactions. At low volumes that is a minor issue. As agentic commerce scales, it becomes a meaningful revenue problem.
The second gap is orchestration. If Visa concentrates merchant-facing agentic infrastructure under one platform, it could capture AI-driven payment volume across other networks, weakening the incentive for merchants to build direct integrations with Mastercard, Stripe, or Google protocols. Merchants that rely on direct point-to-point integrations with individual processors are not positioned to adapt quickly as the protocol landscape continues to shift. The ones who will move fastest are those whose orchestration layer sits above individual processors and can route dynamically as new protocols and payment paths emerge.
Three things merchants should be doing now
The broader rollout of Intelligent Commerce Connect is scheduled for later in 2026. That window is narrow but real.
The merchants best positioned to capture agentic commerce volume are not those with the most sophisticated AI strategy. They are those whose payment infrastructure can handle what is coming without requiring a rebuild.
DEUNA connects merchants to 400+ global payment and fraud providers through a single integration, with orchestration logic that adapts dynamically to transaction context, whether the initiating party is a human or an AI agent. Athia, DEUNA's payments intelligence engine, continuously analyzes transaction-level signals to optimize routing and risk decisions in real time, including for the transaction profiles that agentic commerce produces.
Visa built the rail. The merchants who have the right infrastructure underneath them will be the ones who use it.
