The processor layer is the part of the payments stack nobody talks about. Visa and Mastercard get the headlines. But FIS completing a $13.5 billion acquisition and immediately launching agentic commerce tools tells you where agent transactions will actually be authorised, scored, and settled.

When Visa announces an agentic commerce initiative, it trends on LinkedIn. When Mastercard launches Agent Pay, the fintech press writes it up within hours. When FIS completed the largest issuing acquisition in payments history and launched a bank-facing agentic commerce platform on the same day, the coverage lasted about a news cycle.

That tells you something about how the industry assigns attention. It also tells you where the opportunity is being underestimated.

On January 12, 2026, FIS closed its $13.5 billion acquisition of Global Payments' Issuer Solutions business, formerly known as TSYS. Enterprise value: $13.5 billion. Net consideration: approximately $12 billion. And with the ink barely dry, FIS launched the payments industry's first offering enabling banks to conduct agentic commerce.

The scale of what FIS absorbed is worth stating plainly. TSYS processes more than 40 billion transactions annually across 150+ financial institutions and corporates in 75 countries. It is, by volume, the largest credit processing operation on the planet. FinTech Magazine reported on the scope of the integration: credit processing, fraud detection, loyalty programmes, and a full suite of value-added services.

What FIS already had: debit processing, network services, accounts payable and receivable capabilities. What it now has: the end-to-end issuing stack, rebranded as FIS Total Issuing Solutions.

The strategic signal is in what FIS simultaneously divested. It sold its remaining minority stake in Worldpay to Global Payments. Read that move carefully. FIS exited the merchant acquiring side of payments entirely. It went all-in on the issuer side, the side that says yes or no to every transaction.

In a world where AI agents are about to start initiating purchases, that "yes or no" becomes the most important decision in the stack.

Here is the part of the agentic commerce conversation that keeps getting skipped.

Visa's Intelligent Commerce framework and Mastercard's Agent Pay define how AI agents should identify themselves, request authorisation, and carry credentials. These are protocol-layer specifications. They describe what should happen. They do not make it happen.

The making-it-happen part sits in the processor layer. When an AI agent, acting as a personal digital assistant, sources a product, negotiates a price, selects a pre-approved payment method, and initiates a purchase, that transaction does not flow directly from the agent to Visa. It flows through a processor. The processor runs the authorisation logic. The processor scores the fraud risk. The processor routes the transaction to the correct issuing bank. The processor settles the funds.

As we covered in our analysis of the Q1 2026 agentic commerce stack, the protocol layer has been moving fast. But protocols without processing infrastructure are specifications without engines. FIS is building the engine.

The agentic commerce offering FIS shipped on January 12 introduces a framework called Know Your Agent, or KYA. The concept parallels the familiar Know Your Customer and Know Your Business frameworks, applied to machine-initiated transactions.

The initial use cases are specific: transaction authorisation, fraud detection, loyalty programme integration, and customer servicing. According to the FIS press release, the platform enables banks to identify, authenticate, and authorise transactions initiated by AI agents using pre-approved payment methods and spending mandates.

Both Visa and Mastercard are strategic partners in the launch. That is worth pausing on. FIS is not picking a side in the network protocol competition. It is integrating both, exactly as Fiserv did when it adopted both Visa's Trusted Agent Protocol and Mastercard's Agent Pay for its merchant-side processing. The processor layer is protocol-agnostic by design. It has to be. Banks issue cards on both networks.

The AI functions as a personal digital assistant for the cardholder. It sources products, negotiates terms, and completes purchases using payment methods the cardholder has pre-approved, all within boundaries the bank has verified through KYA. FIS planned availability by the end of March 2026, which means the infrastructure should be live or going live imminently.

McKinsey forecasts that agentic commerce could generate $1 trillion in US retail revenue by 2030, with global estimates ranging from $3 to $5 trillion. Those are protocol-layer numbers. The revenue only materialises if the processing infrastructure can handle agent-initiated transactions at scale.

Consider what happens when an AI agent tries to buy something today. The agent selects a product and a payment method. The transaction hits the processor. The processor's fraud system sees a transaction pattern it has never encountered: no browser fingerprint, no device ID, no human behavioural signals. Without agent-aware processing, the transaction looks suspicious. Approval rates drop. The agent economy stalls before it starts.

That is the problem KYA solves. And it is a problem that can only be solved at the processor level, not at the protocol level and not at the merchant level.

As we explored in our primer on agentic payments, the fundamental challenge of agent-initiated commerce is not discovery or checkout. It is trust. How does a bank know that an agent is acting on behalf of a legitimate cardholder, within approved spending limits, for a genuine purchase? The processor sits at the exact point in the stack where that trust question gets answered.

FIS is not building in isolation. The processor layer is consolidating around agentic commerce across the board.

Fiserv adopted both Visa's Trusted Agent Protocol and Mastercard's Agent Pay Acceptance Framework, positioning to serve its merchant acquiring clients with agent-ready checkout infrastructure. But Fiserv's play is on the acceptance side. FIS is on the issuing side. They are solving different halves of the same problem.

Mastercard has been expanding Agent Pay into Latin American markets, building out regional infrastructure for agent-initiated payments. Visa has over 100 partners building within its Intelligent Commerce sandbox.

But look at where the network initiatives end and the processor work begins. Visa and Mastercard define how agents should present credentials. FIS and Fiserv determine whether those credentials result in approved transactions. The networks set the rules. The processors enforce them.

The parallel to previous payment technology transitions is instructive. EMV chip cards were a network specification. Contactless was a network specification. Both became real when processors integrated them into their authorisation and settlement infrastructure, and merchants inherited the capability without making individual integration decisions. Agent commerce is following the same pattern. The specifications exist. The processor integration is what makes them functional.

The payments industry is spending enormous energy debating which protocol will win the agentic commerce era. Visa Intelligent Commerce or Mastercard Agent Pay? Google's Universal Commerce Protocol or Stripe's machine-readable checkout? Open standards or proprietary frameworks?

Those are interesting questions. But they may be the wrong ones.

The more consequential question is whether the processing layer, the infrastructure that actually authorises, scores, and settles transactions, is being rebuilt fast enough for agent-initiated commerce. FIS absorbing the world's largest issuing business and immediately pointing it at agentic commerce suggests the answer is starting to be yes. But one processor, however large, does not make a market.

If McKinsey's $1 trillion US forecast is even directionally correct, every major processor will need agent-aware authorisation, agent-specific fraud models, and agent-compatible settlement flows. The protocol layer is necessary but not sufficient. The processor layer is where agentic commerce either works or does not.

FIS just placed $13.5 billion on that thesis. The rest of the processing industry will need to decide whether to follow.

If the protocol layer is the specification and the processor layer is the engine, which one does agentic commerce actually depend on more?