Have Payments Cashed in on AI, Or Is It All Still Hype?
AI is everywhere in payments, but the ROI is still uncertain. We asked top execs where the hype finally turns into savings as AI takes the wheel.
To start, what are some real ways AI will actually lower costs in the payments space? Has it?
Claurelle Rakipovic, Chief Product Officer at Pipe: The ROI shows up first where payments companies have the biggest hidden operational drags or labor costs: fraud ops, disputes, onboarding, and reconciliation. It starts with AI that augments humans, not replaces them.
Dan Pinto, CEO & Co-Founder at Fingerprint: Getting rid of legacy fraud prevention systems isn’t realistic in most cases. Instead, companies should layer in AI and newer technologies like device intelligence that can enable them to make more accurate fraud decisions quickly. The real cost savings come from reducing hours spent manually reviewing thousands of transactions and false positives that block legitimate customers. Companies that successfully implement AI and newer technologies can see reductions in operational costs, but only if they implement them thoughtfully.
Mark Sundt, CTO at Stax Payments: The payments industry has quickly adopted AI for a wide range of purposes, but as an industry, we’re still waiting to see the full return on investment. Companies have leaned on large, general-purpose models to scale tools faster. These systems improve efficiency but don’t do much to simplify the daily realities of this complex space. The real ROI for AI will come from smaller, specialized models built to solve specific, high-value problems.
What are the major AI blindspots in payments, and what should companies be thinking of now that they may not be?
Dan Pinto, CEO & Co-Founder at Fingerprint: The biggest blind spot is not having adaptive systems that can handle AI-powered fraud. Combining behavioral analysis with device intelligence is critical. Without comprehensive data on user patterns and device characteristics, even the best models can’t make the right decisions at the speed of digital transactions.
Claurelle Rakipovic, Chief Product Officer at Pipe: When it comes to AI in payments, companies tend to stumble in two big ways: the classic traps and the organizational gaps.
Classic traps are the obvious mistakes. Some firms waste time using AI for jobs that rules already handle better. For example, parsing invoice due dates. Another trap is skipping explainability: if a system declines a transaction but can’t explain why, customers lose trust and regulators raise red flags. A third is underestimating fraudsters, who adapt as quickly as you patch. Without drift monitoring, human review thresholds, and safe fallback rules, you end up playing whack-a-mole. And finally, there’s latency. Payments flows happen in milliseconds. If your AI model is slow or too costly, it’s worse than useless.
Organizational gaps are less visible but just as damaging. Many teams take a vague “let’s add AI somewhere” approach, instead of mapping which real tasks should be automated. Knowledge management is often neglected, too. If information is scattered across CRM, support tools, and docs, the AI can’t pull it together and starts making things up.
AI delivers in payments when it’s aimed at ambiguous, high-value problems, paired with clear explanations and real-time performance. It compounds over time when organizations treat AI not as magic, but as a skillset everyone builds into their daily workflows.
Let’s talk agentic payments. If AI agents take more initiative on behalf of users—say moving money, disputing charges, or adjusting settings—how do we preserve user trust without drowning them in confirmation prompts?
Dan Pinto, CEO & Co-Founder at Fingerprint: The key is setting up smart rules that analyze device signals and user behavior to determine when additional verification steps are actually needed.
For example, instead of requiring all users to complete MFA challenges and other confirmations, assess whether a user is on their trusted device, exhibiting normal behavioral patterns, and operating from expected locations. If the answer is yes, then it makes sense to reduce confirmation prompts (and in certain instances, even eliminate them altogether) to provide a smoother experience for legitimate users. If the answer is no, then keep the extra challenges in place.
This approach maintains account security while allowing legitimate users to complete routine transactions seamlessly.
Mark Sundt, CTO at Stax Payments: As AI systems take on more autonomous roles, the challenge isn’t simply enabling transactions—it’s maintaining trust without creating friction. Payments already rely on confidence and security; when an agent acts on a consumer’s behalf, that foundation must be unshakable. The payments layer should rest on clear guardrails: transparency in data use, auditable governance, and security built into the core rather than added later. Consumers shouldn’t have to question an agent’s actions—they should know the system is operating under strict, visible protections.
As agentic AI scales across payments, which players are best positioned to benefit?
Claurelle Rakipovic, Chief Product Officer at Pipe: The winners in agentic AI will be those with scale, trust, unique data, and who use outcomes as the measure of success (not activity). It’s true for processors, banks, scaled merchants, and vertical focused players.
AI raises the ceiling here for all players. Imagine what it would feel like to have a semi-autonomous CFO for every micro-business. Predicting cash shortfalls and making recommendations before a founder even opens their laptop. That’s transformative for small businesses who’ve never had access to that level of financial tooling.
Mark Sundt, CTO at Stax Payments: Players with both speed and control will benefit most. Fintechs will experiment fastest, offering near-invisible agentic experiences. Banks bring compliance and data strength but move slowly. Processors that own more of the payment stack can embed specialized models directly into transaction flows—reducing fraud, improving interchange, and automating disputes.
AI can make payments fade into the background, but trust, identity, and intent must always remain visible.
Dan Pinto, CEO & Co-Founder at Fingerprint: Fintechs and processors are best positioned because they’re typically less burdened by legacy systems, and can implement AI solutions faster than traditional banks. They can quickly deploy AI for real-time fraud detection, risk assessment, and automated decision-making across millions of transactions daily. Traditional banks that rely only on outdated rule-based systems are most at risk.
Dan Pinto, CEO & co-founder, Fingerprint