On April 8, 2026, Visa launched Intelligent Commerce Connect, a platform that lets AI agents browse, select, and pay for goods autonomously across multiple networks. Combined with Mastercard Agent Pay and agent payment protocols from OpenAI and Google, agentic commerce has moved from concept to production. Merchants that do not prepare risk invisibility in AI-driven purchasing.
What Just Changed in Online Commerce
For most of the last 20 years, the default assumption behind an online transaction has been a human clicking buttons: a consumer browses, compares, and clicks buy. That assumption is now formally breaking.
In a press release on April 8, 2026, Visa unveiled Intelligent Commerce Connect as what it describes as a network, protocol, and token vault agnostic on-ramp to agentic commerce for agent builders, merchants, and enablers. Through a single integration via the Visa Acceptance Platform, it enables secure payment initiation, tokenization, spend controls, and authentication for AI agents. The platform is currently in pilot with Aldar, AWS, Diddo, Highnote, Mesh, Payabli, and Sumvin, with general availability targeted for June 2026.
In parallel, Mastercard Agent Pay has been integrating with Microsoft Copilot Checkout and OpenAI's Instant Checkout program inside ChatGPT. Coverage from Axios described the moment as the major card networks racing to become the payment rail for the agentic economy.
AI shopping agents are not speculative. They are already active inside ChatGPT, Google Gemini, Microsoft Copilot, and Perplexity, completing real purchases for real consumers. According to Digital Commerce 360, AI-referred traffic to US retail sites grew 805% year over year on Black Friday 2025. The infrastructure question has shifted from whether agents will transact to how merchants accept them.
Our take: The Visa launch matters less for what it enables today than for what it signals. When the largest card network sets up a protocol-agnostic on-ramp and names eight pilot partners, it is telling the market that agentic checkout is about to become a standard capability, not a science project. The window to prepare is short.
The Protocol Wars Behind the Scenes
Agentic commerce requires that three historically separate things coordinate: the agent itself, the merchant, and the payment network. Without shared standards, that coordination is impossible at scale. A lot of the work of the last six months has been about settling those standards.
Trusted Agent Protocol. Provides verifiable identity for agents so merchants can tell a legitimate agent from a bot or a fraudster.
Machine Payments Protocol (MPP). Formalizes how autonomous systems authorize and settle payments, including spend limits and intent verification.
Agentic Commerce Protocol (ACP). Originated by OpenAI. Defines how agents discover products, compare them, and execute transactions on behalf of a user.
Universal Commerce Protocol (UCP). Another standard that merchants can adopt to expose inventory and accept agent-initiated purchases.
Agent Payments Protocol / Agent2Agent Protocol. Google's standards for agent identity and payment flows.
Visa Intelligent Commerce Connect's selling point is that it speaks all of the above. Merchants who integrate once can accept agent payments coming through ChatGPT, Gemini, Copilot, Perplexity, and whatever comes next, without individually building to each. That is how this transition avoids the fragmentation that plagued early mobile payments.
For a primer on how these agents are built and why they are becoming more capable, we covered the underlying technology in AI Agents Explained and Multi-Agent Systems Explained.
Why This Is Different From the AI Search Shift
We wrote last year about how AI Overviews are changing how customers find your business. Agentic commerce is the next layer up: it is not just about being findable inside an AI answer, it is about being buyable inside an AI workflow.
The distinction matters operationally. AI search optimization is mostly a marketing problem: structured content, authoritative sources, schema markup. Agentic commerce is a full-stack problem that touches product data, checkout infrastructure, fraud controls, and customer service. Getting cited by ChatGPT does not automatically mean an AI agent can complete a purchase at your store. Those are two different integrations.
Our take: Think of AI search visibility and agentic commerce readiness as two distinct capabilities, both required. Companies that invested in one over the last year have only done half the work.
The Strategic Implications for Merchants and B2B Sellers
The commercial implications split roughly into three categories, each requiring different preparation.
Retailers selling physical goods. This is the most immediate impact zone. If a consumer asks ChatGPT to find the best wireless noise cancelling headphones under $300, the agent needs structured product data, real-time pricing, inventory availability, and a checkout that can accept an agent-initiated payment. Merchants without these capabilities will simply not show up in the answer or will be excluded from the transaction. The 805% year over year Black Friday traffic figure is a leading indicator; the share is small today but growing rapidly.
B2B sellers and SaaS. B2B procurement has a procurement agent use case that is arguably even more valuable than consumer shopping. An agent that can renew contracts, source components, compare SaaS subscriptions, and handle small-dollar purchases automates a large piece of what purchasing teams do manually today. Sellers who expose their catalogs and pricing to agent protocols will capture those transactions; those who do not will lose pipeline they never see.
Service businesses. Service scheduling, bookings, and recurring transactions are all candidates for agentic flows. A customer asking an agent to reschedule a meeting, book a quarterly tune-up, or renew an insurance policy will expect the agent to complete the task end to end, not hand them a phone number.
What Merchants Should Actually Do Right Now
This is not a call to rebuild your e-commerce stack tomorrow. It is a call to make specific, tractable decisions over the next two quarters.
-
Audit your product data for machine readability. Every SKU should have structured metadata: specifications, compatibility, price history, availability, reviews, and unambiguous identifiers. If a human shopper could not make a decision from your product page without calling support, an agent cannot either.
-
Pick a payments strategy. Decide whether you will accept agent-initiated payments via Visa Intelligent Commerce Connect, Mastercard Agent Pay, or both, and on what conditions. Define spend limits, refund flows, and how disputes are handled when the cardholder was not the person who clicked buy.
-
Adopt at least one agent commerce protocol. Visa's approach is protocol-agnostic, but you still need to expose your inventory and accept agent-initiated orders through an adopted standard. ACP and UCP are the leading candidates today. Pick one, integrate, and iterate.
-
Rewire fraud and authorization. Agent transactions look different from human transactions: no mouse movement, no device fingerprint in the traditional sense, no abandoned cart pattern. Your existing fraud rules will either block legitimate agents or let through adversarial ones. Work with your payment processor now.
-
Train customer service on agent-initiated transactions. When an agent makes a mistaken purchase, the consumer contacts your support team, not the agent vendor. Support teams need scripts, refund authority, and escalation paths specific to agent flows.
The Risks Nobody Is Pricing In Yet
Agentic commerce is not a pure upside story. There are risks to think through before scaling participation.
Attribution collapse. When an agent selects your product from a set of comparable options, why? Was it your brand, your price, your data quality, or a quirk in the agent's ranking logic? Traditional analytics assumes a human path to purchase. Agent-mediated purchases are harder to attribute, which complicates marketing investment decisions.
Price compression. Agents optimize aggressively. If your differentiation depends on brand perception or human persuasion, and the agent cannot see that, you will get commoditized on price faster than you expected.
Dispute exposure. Card networks are still writing the rulebook for chargebacks on agent-initiated purchases. Merchants who move early will be building processes without a complete regulatory and liability framework. Plan for this, do not get surprised by it.
Key Takeaways
- Visa Intelligent Commerce Connect launched April 8, 2026, providing a protocol-agnostic on-ramp for AI agents to initiate payments across multiple networks and major commerce protocols.
- Mastercard Agent Pay and card network integrations with ChatGPT Instant Checkout and Microsoft Copilot Checkout confirm that agentic commerce is being built as standard payment infrastructure, not a niche experiment.
- AI-referred traffic to US retail sites grew 805% year over year on Black Friday 2025, according to Adobe Analytics cited by Digital Commerce 360, a leading indicator of mainstream adoption.
- The key capabilities merchants need are machine-readable product data, support for at least one agent payment protocol, agent-aware fraud controls, and updated customer service playbooks.
- Agentic commerce is distinct from AI search optimization. Companies that invested in one but not the other have done half the work.
The businesses that move early on agentic commerce will have a meaningful advantage. If you want to be one of them, let's start with a conversation.