Payments Execs Say AI Forces New Payments Decisions

The Intelligence Around the Transaction Becomes the New Prize

Payment infrastructure has traditionally created value by executing an instruction accurately and efficiently. Artificial intelligence makes the decision surrounding that instruction more valuable.

“Payment execution alone is increasingly commoditized. The durable value as I see it today is in making the right payment decisions,” Tungsten Automation Head of Payments and Embedded Finance Andrew Ng told PYMNTS.

Large merchants and enterprises frequently operate across multiple processors, payment service providers, geographies and payment methods. Each transaction produces information about authorization rates, routing, costs and fraud. The problem has historically been turning all of that information into action quickly enough to improve the next transaction.

“There’s a ton of payments optimization potential in these payments rails, particularly in complex multi-PSP [payment service provider] environments,” Adam Hiatt, executive vice president of product strategy at Spreedly, told PYMNTS.

The transaction still matters. But the competitive advantage increasingly sits in understanding enough about the transaction to decide how it should happen.

“As an AI agent begins to assist with purchasing decisions, the payment ecosystem has to make that value chain machine readable,” Mike Magennis, senior director of Strategy, Networks at FIS, said. “An agent needs to understand not just whether a payment credential works, but which choice delivers the best relevant outcome for that consumer.”

AI Is Making the Boring Plumbing Strategically Valuable

Of course, intelligence gets tangled in dependencies. AI only reasons over information it can reach. That makes the fragmented architecture built up over decades of financial technology a new constraint on AI deployment.

“There’s tremendous attention right now on what AI can say, but there’s a lot less attention to what AI can actually reach,” Chris Trainor, head of platform strategy, innovation and AI at Paymentus, told PYMNTS.

An AI agent may understand that a customer wants to change a payment date. But customer data can sit in one application, billing rules in another, payment credentials in a third and communications preferences somewhere else. Every additional system creates another integration, permission structure and set of business rules that an agent must navigate.

It’s the same situation on the consumer side too, as it relates to discoverability.

“If you’re not focused on that today, if you’re not prepping your brand to show up in these platforms, if you’re showing up 10th there, but first in traditional search, that’s the place you got to be putting your priorities,” Mike Storiale, senior vice president, AI Technology and Transformation at Synchrony, told PYMNTS.

“If the SKUs aren’t available and findable, nothing else matters,” Spreedly’s Hiatt said.

After all, the technologies with the greatest potential for AI-driven optimization are the less glamorous infrastructure pieces of the ecosystem. That includes common commerce protocols, financing options, authentication, proof of intent and ultimately a new trust layer for transactions initiated by machines.

“My first pick is the global compliance piece,” Dave Ruda, vice president, product, at Billtrust, told PYMNTS, adding that AI may end up increasing the value of the plumbing it is supposedly leapfrogging. “If you can’t get that done, then you’re not going to make it to the second round.”

The Wallet Isn’t the Hard Part; Permission Is

Once machines can understand the financial environment, the next problem is allowing them to act inside it.

“Authentication remains foundational in an AI landscape,” Trainor said. “The mistake that many have is expecting that model itself to become the system of record or a transaction processing engine. The future is not unconstrained artificial intelligence. It is governed agency.”

“There are good bots, and there are bad bots,” Hiatt said. “And so, KYA, the know your agents model, that idea as a business is a new category.”

The answer may also lie in the information, aka the intelligence, that comes with a payment. Structured invoice information can give an AI system context about what buyers are purchasing, who receives payment and why a transaction exists. Combine that with payment history, counterparty information, policies and payment-rail data, and AI can potentially help determine transaction execution while improving reconciliation and compliance.

“This builds a solid substrate for agentic payments that no one is talking about in that language,” Tungsten Automation’s Ng said.

AI Is Moving the Economic Battleground Away From the Checkout

Technologies such as Model Context Protocol can allow AI applications to interact with external tools and information. For financial software, the significance isn’t simply better integration. An agent can potentially interact directly with the underlying system rather than requiring an employee, or a consumer shopper, to navigate its interface.

“You don’t need front end to do anything. You just need to be able to talk to the agent,” Ruda said, describing the emerging model as “headless” financial software.

“Think about all the tasks you do now being spun up by agents and being automatically executed,” Ruda added. “That’s pretty powerful and very real.”

A consumer could, for example, authorize an agent to reorder household goods below a certain amount or book travel within defined parameters without approving each transaction individually.

FIS’ Magennis said he would initially favor “more bounded use cases where there’s more customer permission-focused deployments, where there’s spending limits that are clear, where accountability is clear.”

But that seemingly simple instruction requires infrastructure that can preserve the relationship between identity, intent, merchant, credential, spending limit and transaction.

Synchrony Senior Vice President, AI Technology and Transformation Mike Storiale sees the result as something larger than authentication.

“You’re going to need intent, you’re going to need authentication,” he told PYMNTS. “But so many of the building blocks to get to that trust point have to be there first.”

A bank, network, issuer, BNPL provider or merchant that waits until checkout to influence the transaction may discover that an AI system somewhere else has already made the most consequential decision.

Taken together, the trendlines shared by the six experts independently suggest that the transaction itself is becoming the end of the story rather than the beginning of it.

AI may finally create the ROI case for years of unglamorous payments modernization.