How Visa's Intelligent Commerce Connect Powers AI Agent Payments

When Your AI Agent Picks Up the Tab

In December 2025, an AI agent purchased a pair of Bose headphones after consulting Consumer Reports — no human clicked "buy." That transaction was one of hundreds of secure agent-initiated purchases completed during Visa's closed beta pilot, a quiet proof-of-concept that signaled a fundamental shift in how commerce might work. Four months later, Visa has taken the next step: a platform designed to make AI-driven buying not just possible, but scalable.

On April 8, 2026, Visa unveiled Intelligent Commerce Connect (ICC), a new solution within the company's broader Intelligent Commerce portfolio. ICC is designed to serve as a network-agnostic, protocol-agnostic, and token vault-agnostic on-ramp to agentic commerce — the emerging model in which AI agents autonomously discover products, compare prices, and complete purchases on behalf of consumers and businesses.

The timing is deliberate. According to McKinsey projections cited by The Letter Two, agentic commerce sales could surpass $5 trillion by 2030. Visa is betting that whoever builds the trusted payment infrastructure underneath that wave will capture outsized value — and it intends to be that builder.

What Intelligent Commerce Connect Actually Does

At its core, ICC is an abstraction layer. Through a single integration via Visa's existing Acceptance Platform, merchants, agent builders, and payment enablers gain access to secure payment initiation, tokenization, spend controls, and authentication for AI-driven transactions. The platform processes both Visa and non-Visa card payments, a distinction worth emphasizing: Visa is not restricting this to its own network.

The platform supports four major agentic commerce protocols, according to Visa's press release:

• Trusted Agent Protocol — Visa's own standard, introduced in October 2025
• Machine Payments Protocol (MPP) — developed by Stripe and Tempo
• Agentic Commerce Protocol (ACP) — from OpenAI
• Universal Commerce Protocol (UCP) — from Google

By supporting all four, Visa is positioning itself as the neutral settlement layer regardless of which protocol wins the standards race. If a merchant integrates with ICC, its inventory becomes discoverable to AI agents regardless of which protocol those agents speak.

For merchants, the value proposition extends beyond payment processing. ICC helps make product inventories — descriptions, specifications, and pricing — accessible within AI platform experiences, so agents can discover, select, and check out without the consumer ever leaving the AI interface. Visa also handles PCI compliance orchestration for enablers supporting merchant transactions, removing a significant technical burden.

"Intelligent Commerce Connect brings that same, trusted payment acceptance infrastructure into the emerging world of AI-driven commerce," said Andrew Torre, President of Value-Added Services at Visa.

From Pilot to Platform: The Road to ICC

ICC did not emerge from a vacuum. Visa has been methodically building toward this moment since at least late 2025.

In its December 2025 announcement, Visa reported completing hundreds of secure agent-initiated transactions across several closed-beta pilots in the United States. The participants illustrate the breadth of use cases Visa envisions:

• Skyfire partnered with Consumer Reports to build a product recommendation agent that completed real purchases — the Bose headphones transaction mentioned above.
• Nekuda enabled fashion purchases through Gensmo's app, buying from retailers like Fabrique and Honeylove.
• PayOS powered agent-driven checkout at luxury watch retailer Jomashop through its BeyondStyle agent.
• Ramp demonstrated B2B bill pay with automatic cashback capture.

The December announcement also revealed the scale of Visa's partner ecosystem: over 100 partners globally, with more than 30 actively building in the Visa Intelligent Commerce sandbox and over 20 agents or enablers integrating with the platform.

That same announcement included a striking data point: 47% of U.S. shoppers already use AI for shopping tasks, according to Visa. While "shopping tasks" covers a broad range of activities — from product research to price comparison — the figure suggests that consumer behavior is already shifting toward AI-assisted commerce, even before autonomous purchasing becomes mainstream.

The current ICC pilot includes seven named partners — Aldar, AWS, Diddo, Highnote, Mesh, Payabli, and Sumvin — with general availability expected in 2026.

The Trust Problem: Why Infrastructure Matters More Than Intelligence

The most sophisticated AI agent in the world is useless for commerce if no one trusts it with their credit card. This is the core challenge Visa is attempting to solve — and it may be more difficult than the AI itself.

Pavan HS, Senior Director of Product for Agentic Commerce at Visa, framed the issue directly in a January 2026 blog post on the Visa Acceptance Solutions site: trust will scale gradually, starting with small transactions before progressing to larger ones over several years. The parallel to earlier payment innovations is apt — consumers did not immediately trust online checkout or contactless payments either.

Visa's approach to the trust problem draws on its existing infrastructure. According to BanklessTimes, the consumer protection model for ICC preserves several familiar guardrails:

• Cardholders can set spending caps and maintain merchant approval lists
• Agent permissions can be reviewed and revoked at any time
• Card issuers retain control over risk rules
• Traditional fraud checks and receipt documentation remain intact
• Transactions appear as standard card payments to merchants

This last point is strategically important. By making agent-initiated transactions look like ordinary card purchases to the merchant's existing infrastructure, Visa eliminates the need for merchants to build entirely new acceptance systems. The complexity of agentic commerce is absorbed by Visa's platform rather than pushed to the merchant.

Tokenization plays a central role in the security architecture. As Visa's Acceptance Solutions blog explains, the platform leverages existing network token technology that merchants already use for traditional digital payments. Agents never handle raw card credentials; instead, they operate with tokenized representations that limit exposure even if a compromised agent is exploited.

The Protocol Wars: Visa's Neutral-Layer Strategy

The agentic commerce landscape in 2026 is characterized by a proliferation of competing protocols, each backed by a major technology or financial services company. Visa's Trusted Agent Protocol competes — and now coexists — with Stripe's Machine Payments Protocol, OpenAI's Agentic Commerce Protocol, and Google's Universal Commerce Protocol. Other major payment networks are developing their own approaches to agentic commerce.

This fragmentation creates a real problem for merchants: which protocol should they support? Integrating with each independently would be expensive and operationally complex.

ICC's value proposition is to collapse that complexity into a single integration point. A merchant connecting to ICC through Visa's Acceptance Platform gains compatibility with all four supported protocols simultaneously, without needing to understand the technical differences between them. Visa handles the translation and orchestration.

This is a familiar playbook for Visa. The company built its original business by sitting between consumers and merchants as a trusted intermediary in a fragmented payments landscape. ICC applies the same logic to a new fragmentation — agentic protocols rather than card networks. The bet is that even as AI agent builders and protocol designers compete for developer mindshare, transactions will still flow through a common settlement layer.

The network-agnostic design — processing non-Visa cards alongside Visa cards — underscores the ambition. Visa is not trying to win the agentic commerce market by restricting it to Visa credentials. It is trying to win by becoming the infrastructure that everyone uses, regardless of card network.

The Scale Advantage: Why Visa Starts Ahead

Visa's pitch for ICC leans heavily on the scale of its existing network. According to Visa's corporate product page, the company processes over 300 billion transactions annually across more than 150 million merchant locations, connecting 4.8 billion payment credentials globally.

These numbers represent a significant structural advantage in the agentic commerce race. An AI agent's utility is directly tied to where it can transact. An agent that can purchase from any of Visa's merchant locations has dramatically broader coverage than one restricted to merchants that have adopted a newer, less established payment protocol.

The B2B capabilities embedded in the Intelligent Commerce suite extend this advantage further. Visa's product page describes automated inventory detection, supplier sourcing, and automatic invoice reconciliation — functions that could make AI agents particularly valuable for procurement and supply chain management, where repetitive, rule-based purchasing decisions are already common.

Visa has signaled a phased global rollout strategy, with North American enterprise adoption leading the way.

What Could Go Wrong

Visa's strategy is not without significant risks.

The stablecoin alternative. While Visa anchors agentic commerce to traditional card rails, stablecoin-based payment systems offer a fundamentally different architecture — one where machine-to-machine transactions settle on blockchain networks without intermediaries. If AI agents gravitate toward stablecoin rails for their lower friction and programmability, Visa's card-centric approach could find itself competing against a structurally cheaper alternative for certain transaction types.

Protocol consolidation uncertainty. ICC's value proposition depends partly on protocol fragmentation persisting long enough for a neutral aggregation layer to be valuable. If one protocol — say, OpenAI's ACP, given the company's massive developer ecosystem — achieves dominant market share, the need for a multi-protocol aggregator diminishes.

Consumer trust at scale. Visa's own internal framing acknowledges that trust will build gradually over several years, starting with low-value transactions. The gap between consumer willingness to let an AI agent handle a small purchase versus a high-value or complex transaction like travel booking remains largely untested. A single high-profile incident of agent misbehavior — unauthorized purchases, data exposure, or fraud — could set adoption back significantly.

Merchant adoption friction. While ICC is designed to be a single integration, merchants must still actively opt in and make their catalogs accessible to AI agents. Catalog standardization is a nontrivial challenge, particularly for businesses with complex or dynamic inventories.

The Bigger Picture: Who Owns the Checkout?

The launch of Intelligent Commerce Connect is ultimately about a deeper strategic question: in an economy where AI agents increasingly initiate purchases, who controls the checkout experience?

Today, that control belongs to merchants and payment processors. But agentic commerce shifts the point of purchase from a merchant's website or store to an AI platform's interface. The consumer may never see the merchant's brand, never browse a product page, never interact with a checkout flow. Instead, the agent handles everything — discovery, comparison, selection, and payment — within its own environment.

This creates both an opportunity and a threat for Visa. The opportunity is that Visa becomes even more essential as the invisible payment layer that enables these transactions. The threat is that if agentic platforms build their own payment infrastructure — bypassing Visa entirely — the company could be disintermediated from a growing share of commerce.

ICC is Visa's answer to that threat: embed itself so deeply into the agentic commerce stack that building around Visa becomes harder than building through it. By supporting multiple protocols, processing multiple card networks, and integrating with the agent builders themselves, Visa is constructing what amounts to a tollbooth on the agentic economy.

Whether that tollbooth becomes as indispensable as Visa's original network remains to be seen. But with general availability expected in 2026 and a partner ecosystem already numbering over 100 companies, Visa is not waiting for the agentic economy to arrive. It is building the roads.

Key Takeaways

• Visa's Intelligent Commerce Connect is a protocol-agnostic platform that lets AI agents complete purchases across multiple card networks through a single merchant integration, currently in pilot with seven named partners and expected to be generally available in 2026.

• Four major agentic protocols supported — Visa's Trusted Agent Protocol, Stripe/Tempo's MPP, OpenAI's ACP, and Google's UCP — reflecting a strategy of neutral aggregation rather than proprietary lock-in.

• Consumer trust remains the key bottleneck. Visa's own experts anticipate a gradual adoption curve starting with low-value transactions before progressing to larger purchases over several years.

• The competitive landscape is crowded. Competing payment networks, stablecoin-based payment rails, and the possibility of AI platforms building their own payment infrastructure all represent alternative paths that could limit Visa's centrality.

• Scale is Visa's strongest moat. With 4.8 billion credentials, over 150 million merchant locations, and 300 billion-plus annual transactions, ICC starts with coverage that no competing agentic payment solution can match today.