Payments CEOs Weigh In on How Much Autonomy Is Too Much

Now picture the last time a payment didn’t work. You remember that one. Because failure is rarely invisible when money is involved.

In 2025, that contrast powered the most candid conversations in PYMNTS’ “What’s Next in Payments” series. Month by month, we convened PYMNTS clients and industry leaders across banking, payments, FinTech and the digital economy to talk about what changes when commerce’s infrastructure gets smarter, faster and more autonomous.

These aren’t abstract trend forecasts. They’re operator decisions made inside regulated environments, under fraud pressure, and on timelines set by real-time commerce. The series is grounded in PYMNTS Intelligence’s broader data-and-analysis work focused on “what’s now and what’s next” in payments, commerce and the digital economy.

If 2024 was the year of AI pilots, 2025 was the year of AI governance. Our interviews showed a sharp shift from “Can we?” to “How do we do this without blowing up risk, compliance or customer trust?” In January’s gen AI conversations, Stax CTO Mark Sundt captured the balancing act succinctly: “Governance, cost and utility form the three points of the [gen AI] triangle we’re trying to balance.”

By July, the conversation had escalated from copilots to agents — software that can make decisions, initiate workflows and collaborate across systems. i2c CEO Amir Wain told us: “We now actually think of agents as one of the boxes in our org chart.” If AI gets a job title, it also needs controls like permissioning, logging, explainability and human override because “autonomy” in payments is another word for “risk.”

Theme 1: Agentic AI

Across the spring and summer installments, executives repeatedly described the back office as agentic AI’s proving ground: onboarding, underwriting, payment orchestration, dispute handling and compliance monitoring. The point wasn’t to eliminate people. It was to eliminate waiting — manual queues, swivel-chair processes, and brittle workflows that can’t scale with always-on commerce.

But the interviews also converged on a warning: Agentic AI is only as effective as the infrastructure beneath it. Bad data doesn’t just produce wrong answers; it can produce wrong actions at machine speed.

Theme 2: Data collaboration

The industry’s other obsession in 2025 was not simply “more data,” but “more reliable signals.” In September’s banking data-deluge conversations, fraud leaders described a layered approach: real-time behavioral insights, historical context, and — in carefully bounded ways — consortium intelligence that helps the ecosystem raise the bar.

Entersekt Chief Product Officer Pradheep Sampath summarized the tightrope in one of the year’s most provocative lines: “Where the industry needs to go forward is to share data in a responsible manner across consortia, while preserving fair competition.”

Theme 3: CEOs treat data as power, not exhaust

By November’s “Year of the CEO” installment, the narrative widened. Leaders talked about data discipline as a leadership mandate: the prerequisite for AI-driven automation, identity-centric trust, and real-time decisioning. Boost CEO Dean Leavitt delivered one of the most practical quotes of the year: “If you trust AI blindly, you do so at your own peril. But if you ignore it … you do so at your own peril.” The takeaway: the competitive advantage is not “more AI.” It’s knowing where AI belongs, and where it doesn’t.

Theme 4: Trust

Identity and resilience showed up again and again — not as compliance checkboxes, but as market differentiators. In the fall, leaders told us identity is becoming strategic: “smart friction,” interoperable credentials, and architectures that let the right users in quickly without letting attackers in at all.

December’s “Unsung Heroes” edition made the same point in operational terms: uptime is now part of the product. Entersekt Chief Strategy Officer Dewald Nolte put it with Black Friday clarity: “We had 100% uptime this year. What you want is a Black Friday and not a blackout Friday.” Payments outages aren’t IT incidents anymore; they’re trust events.

Theme 5: Growth shifts from rails to the layer around the rails

Even when AI dominated the conversation, a quieter business-model shift kept surfacing: margins are compressing, and “processing” alone isn’t enough. In the value-added services installment, execs described an industry racing to bundle fraud tools, analytics, loyalty, installments, orchestration and open-banking rails into every transaction — turning payments from a utility into a platform.

Theme 6: Gen Z

The consumerization of expectations is now flowing into B2B as Gen Z climbs into decision-making roles. In June’s Gen Z discussions, i2c’s David Durovy set the stakes bluntly: “Consumer cohorts splinter faster than ever before. If your infrastructure can’t pivot with them, you’re done.” The message wasn’t about marketing. It was about architecture.

AI was the headline because it’s loud and legible. But the interviews made another point that’s impossible to ignore: AI doesn’t replace fundamentals, it punishes weak ones. Clean, governed data. Modular infrastructure. Identity that scales. Resilience that stays boring. And value propositions that treat payments as a product experience, not a back-end event.

If 2025 was the year payments leaders stopped asking whether AI matters, 2026 will be the year they prove where it belongs: inside auditable workflows, inside trusted identity layers, and inside systems built to keep money moving when everything else gets messy. “What’s Next in Payments” will keep tracking that shift one month, one argument, and one hard quote at a time.