
i2c's Matthew Pearce explains why payment processors must now compete on real-time AI decisions, not just uptime and cost, and why fraud precision beats severity.
The old competition between payment processors – uptime, scale, per-transaction cost – is giving way to something harder. Institutions now expect partners that can read transaction data in real time, cut fraud without killing approvals, and adapt rules faster than criminals shift tactics. Reliability is still the floor. Speed and intelligence are the new ceiling.
Matthew Pearce, vice president of fraud risk management and dispute operations at i2c, described the change in a recent conversation. The conversations he hears from financial institutions used to center on reliability and operating costs. Now they center on execution speed and adaptability. “If the customer experiences too much friction, that's painful for them, and they put you at the back of the wallet,” he said. The same happens if they experience fraud. That captures the central dilemma.
The objective is precision, not severity. Introduce friction where the evidence warrants. Let legitimate transactions flow. Conversion rates, fraud losses and customer loyalty become interconnected measures, not separate performance indicators. The balance between fraud prevention and customer experience is becoming the defining challenge.
Pearce cited research from LexisNexis showing that financial institutions incur more than $5 in associated costs for every dollar directly lost to fraud. The operational expenses often exceed the losses themselves. Yet cutting those costs by tightening controls can push good customers away. Relaxed controls preserve approvals but may increase both losses and dissatisfaction.
Architecture matters. Fragmented systems create blind spots that limit what AI can see. Unified platforms allow information to flow across fraud, disputes and risk management functions, Pearce said. The broader industry is moving in the same direction.
Static rules have a short shelf life. Routine decisions can be automated through established logic. Adaptive systems that learn continuously from new transaction patterns are becoming essential. Pearce said i2c combines data scientists and fraud analysts in ongoing feedback cycles designed to refine models while keeping approval rates high. The goal is to reduce false positives without sacrificing fraud prevention. Periodic model updates cannot keep pace with criminals who adopt new techniques rapidly, he added.
The next frontier is agentic AI. The hardest problem may not be authenticating the credential but determining intent. “What is the intent of the purchase?” Pearce asked. “Did the customer really intend to buy that?” Credentials can be valid, tokens authentic, while the authority behind a transaction remains uncertain. Authentication alone will not resolve every dispute. Fraudsters are already using AI aggressively. Defenses must be equally sophisticated.
Transparency becomes a governance issue. Organizations need to understand and explain how AI reaches decisions, particularly when automated systems affect approvals, declines or account actions that touch consumers directly. Those questions will become more pressing as AI agents participate more actively in commerce.
Pearce expects processors to contribute intelligence alongside infrastructure by the end of the decade. Real-time product configuration, continuously updated fraud controls, dynamic personalization – these may become baseline expectations rather than premium add-ons. Shared behavioral insights and explainable models could prove as valuable as transaction processing itself. “Speed, precision and transparency have to rise together,” he said. “If any one of those slip, the process really goes off the track.”
Prepared with AlphaScala research tooling and grounded in primary market data: live prices, fundamentals, SEC filings, hedge-fund holdings, and insider activity. Each story is checked against AlphaScala publishing rules before release. Educational coverage, not personalized advice.