On June 16, Adyen announced Adyen Agentic, a three-layer stack for letting AI agents shop and pay on a merchant's infrastructure. The launch partners were not a startup roster. They were Visa, Mastercard, American Express, and Salesforce. The stock rose about 4 percent the same day. For a company that processes payments for Spotify, Uber, and Meta, that is the market deciding agentic commerce is now a revenue story rather than a conference panel.

We have spent months tracking the protocols that are supposed to govern agent payments. Adyen just did something more interesting than pick one. It built a layer that sits above all of them and sells itself as the translator between them.

Adyen is not betting on which agent payment standard wins. It is betting that none of them does, and positioning to be the layer every merchant routes through in the meantime.

That is either the most pragmatic read of the moment or the clearest sign that the standards fight is about to be settled by infrastructure rather than committees. We think it is both, and that the implications run deeper than the launch coverage suggested.

What Adyen actually shipped

Adyen Agentic has three layers, and the names matter less than what each one claims.

The first layer handles discovery, the work of making a merchant's catalog and policies legible to an agent before any purchase happens. The second handles the cart, the construction and validation of an order an agent assembles on a shopper's behalf. The third is the one to watch. Adyen calls it the agentic payments layer, and according to Adyen, it covers authentication, token portability, merchant-of-record preservation, and risk management "across evolving protocols."

Read that last phrase slowly. Adyen is not committing to Google's Agent Payments Protocol, or to the Agentic Commerce Protocol, or to Visa's agent token work. It is promising to keep a merchant's tokenization and merchant-of-record position intact no matter which of those wins, loses, or merges. As PYMNTS reported, the pitch to enterprise merchants is continuity: plug in once, and Adyen absorbs the churn underneath.

Merchant of record is the quiet center of this. The merchant of record is the entity that legally sells to the customer, owns the transaction, carries the chargeback liability, and keeps the tax and compliance obligations. In card commerce, that role is settled. In agent commerce, it is not. When an agent buys a pair of running shoes for you, is the merchant of record the shoe brand, the agent's operator, or the platform that ran the checkout? Adyen's answer is that it should stay the merchant, and that Adyen will do the work to keep it there.

The case that this pushes agentic commerce forward

Start with the honest version, because it is strong.

For two years the agentic commerce story has been demos and specifications. We have written before about how the payment rails moved inside the agent, and about the growing pile of competing standards in the agentic commerce protocol map. What has been missing is a processor of consequence putting production infrastructure behind it, with the networks attached.

Adyen Agentic is that. It is not a sandbox. It is a commercial product aimed at the enterprise merchants Adyen already serves, with Visa, Mastercard, and Amex willing to attach their names. That changes the conversation for a merchant evaluating whether to support agent traffic. The question moves from "is this real" to "do I route it through Adyen or build it myself."

The 4 percent move tells you the second part. Investors do not reprice a payments company on a press release about a protocol. They reprice it when they believe a new transaction stream is coming and that the company has a defensible place to stand in it. A processor that becomes the default agent-checkout layer for its merchant base earns fees on a category that did not exist last year. The market read Adyen Agentic as a moat, not a science project, and the stock followed.

There is a structural reason the bet is credible. Adyen runs both sides of the transaction, acquiring for merchants and connecting to the networks. That two-sided position is exactly what an agent checkout needs, because the agent's request has to be authenticated, authorized, and settled in one flow. A pure protocol cannot do that. A processor can.

The case that it slows the standards down

Now the uncomfortable version.

A translator makes money when translation is necessary. If one agent payment standard won outright, the value of a layer that abstracts away the differences would fall toward zero. Adyen's product is most valuable in exactly the world where AP2, the Agentic Commerce Protocol, and Visa's token work all persist, none dominant, each with enough merchant and platform support to matter. Adyen has built a business that is long fragmentation.

That is not a criticism of Adyen. It is a rational response to genuine uncertainty, and merchants will be glad to have it. But it means one of the most capable players in payments now has a commercial incentive for the standards question to stay open. Every quarter the protocols compete without resolving, the translator layer gets more indispensable.

This is how infrastructure quietly decides outcomes that committees were supposed to. The W3C and the standards bodies can publish all the specifications they like. If the practical path for a merchant is to hand the whole problem to a processor that promises to handle "evolving protocols," the specifications stop being the thing merchants integrate against. They become implementation details inside Adyen. The standard that wins, in practice, is the abstraction layer, not any of the protocols underneath it.

We have seen this movie in payments before. Tokenization was supposed to be an open capability. It became, for most merchants, a feature of whichever gateway they already used. The merchant never touched the underlying network token directly. The gateway owned the relationship. Adyen is making the same move one layer up, for agents.

The gap underneath: translation is not verification

Here is what the launch does not solve, and it is the part that should worry a merchant more than protocol churn.

Adyen Agentic moves tokens, preserves merchant of record, and manages risk. None of that answers the question at the heart of an agent payment: did the user actually authorize this purchase, and is that authorization still valid at the moment the charge clears? Translation is not verification. A layer can faithfully convert an agent's request from one protocol's format to another and still have no idea whether the human behind the agent ever agreed to it, or revoked the agent's permission five minutes ago.

This is the gap we keep returning to. Google shipped the closest thing the industry has to a portable answer, which we covered in the only agent payment mandate that travels. A mandate is a signed instruction from the user that says what the agent may buy, from whom, and up to what limit. The promise of a mandate is that any party in the chain, including a merchant with no relationship to the user's issuer, can verify it. That is the missing primitive. Adyen's announcement gestures at authentication, but a processor authenticating its own connection to an agent is not the same as a merchant verifying that the user's mandate is genuine and live.

The risk for merchants is subtle. A translator that smooths over protocol differences can also smooth over the absence of real authorization, because everything in the flow looks well-formed. The order is valid, the token is portable, the merchant of record is preserved, and the charge clears. Whether the user said yes is a separate fact that nothing in the stack is required to prove. Fraud in agent commerce will not look like a malformed request. It will look like a perfectly formatted purchase the customer never approved.

Who wins, who loses

Adyen wins first and most clearly. It converts an existential question, will agents disintermediate processors, into a product line where it sits in the middle of agent checkout and earns fees. Its enterprise merchants win convenience and a single throat to choke, at the cost of deeper dependence on a layer they do not control.

The networks are the interesting case. Visa, Mastercard, and Amex attached their names, which reads as endorsement. It is also a hedge. The networks have their own agent ambitions, including token frameworks and identity work, and we have written about how their own agentic pilots are accelerating. By partnering with Adyen rather than competing on the checkout layer, they keep their tokens flowing through agent transactions without owning the merchant-facing experience. That is a reasonable trade today. It also cedes the customer-facing position in agent commerce to a processor, which the networks spent two decades trying not to do in card commerce.

The losers, if there are any yet, are the protocols themselves as independent standards. Every merchant that adopts a processor-owned translation layer is a merchant that no longer cares which protocol wins. The specifications still matter to the labs and the networks building them. They matter less and less to the merchants who were supposed to adopt them.

What to watch

Three things will tell us how this plays out.

First, whether a network ships its own translator. If Visa or Mastercard decides that abstracting across protocols is too important to leave to a processor, the partnership turns into competition fast, and the merchant suddenly has two universal translators to choose between. That is a different market than the one Adyen launched into.

Second, whether the mandate layer hardens. If a verifiable, portable mandate becomes the norm, the authorization gap closes and the translator becomes a cleaner, safer product. If it does not, the gap sits underneath every agent transaction Adyen routes, waiting for the first large fraud event to expose it.

Third, whether merchants notice the dependency. The convenience of handing agent commerce to a processor is real. So was the convenience of handing tokenization to a gateway, and merchants spent years afterward trying to regain leverage over a relationship they had given away. The merchants who think about that now will negotiate differently than the ones who do not.

Adyen built the smart product for this moment. The moment is one where nobody knows which standard wins, and the safest business is to sell neutrality between them. That is genuine progress for agent commerce, and it is also a bet that the hardest questions stay unanswered long enough to be profitable.

Sources

If the winning move in agentic commerce is to sell neutrality between the protocols, who is left with a reason to make one of them win?

Charlie Major is a Product Development Manager at Mastercard. The views and opinions expressed in Major Matters are his own and do not represent those of Mastercard.