Agentic AI Poses a Risk to Mastercard's Business Model

Chris Skinner, a longtime financial technology commentator, published an interview this week arguing that Agentic AI challenges the card network model. Credit cards were designed for humans, he said. They assume a person authenticates and disputes. An AI agent that makes 20,000 transactions an hour does not fit that architecture.

Skinner sketched a future where the transaction becomes irrelevant. A consumer tells an AI, "Get me to Singapore for the meeting." The AI decides the payment rail. It optimises for success rate and settlement cost. It does not care about Visa or Mastercard branding. "Algorithms are ruthlessly rational," he said. "They don't care about rewards points or metal cards."

This thesis challenges the revenue engine of Mastercard and Visa. Interchange fees and cross-border fees are built on a human-driven purchase flow. If AI agents route payments through stablecoins or account-to-account rails, those fees disappear. Mastercard's Alpha Score sits at 72, reflecting a solid business that faces an open question about its role in an agentic economy. Read the full MA stock analysis.

The card networks are not standing still. Visa and Mastercard are developing "Know Your Agent" frameworks. Worldpay and Trulioo are building a "Digital Agent Passport." Skinner called these rational first steps. "Cards solve the distribution problem," he said. He warned that identity frameworks designed for static human accounts do not fit agents that can create and destroy sub-agents in seconds.

The deeper risk, Skinner argued, is that the payment rail becomes invisible. An AI does not choose a card. It chooses an orchestration layer that provides identity, authentication, fraud detection and settlement certainty. Account-to-account payments settle in real time. Stablecoins cross borders without correspondent banking. An AI optimising for cost and certainty will choose the rail that works, not the one with the best brand.

"The winners won't be those with the fastest payment rails," Skinner said. "They'll be the ones that can build an identity and trust infrastructure robust enough." If Mastercard and Visa successfully reposition as trust orchestration platforms, the risk recedes. If they fail to adapt, software agents will default to infrastructure built for machines, not humans.

Skinner raised a more uncomfortable possibility. AI agents will become consumers in their own right. They will buy services from other agents. A travel AI hires a visa-processing AI. An energy AI buys weather forecasts. Machine-to-machine commerce, he said, may become larger than human commerce because software transacts millions of times a second over tiny amounts. That economy barely involves cards.

The fraud threat also escalates. Skinner described autonomous fraud agents that spend weeks learning a company's behaviour before making a move. They test payment limits with tiny transactions. They clone voices from podcasts. Defence, he said, becomes autonomous too. Banks deploy defensive AI agents. "Bot-to-bot wars," he called it. The card networks' fraud detection services must evolve to handle attacks that never involve a human.

Skinner predicts 35% of routine consumer payments will be initiated by AI by 2030. The inflection point is whether consumers trust their AI to spend on their behalf. The UK's National Payments Vision and the FCA's AI Live Testing programme suggest regulators are preparing for this shift.

Mastercard's tokenisation and authentication services face a test. The decision-maker is no longer a person. It is an algorithm. Skinner's timeline gives the industry roughly five years to prove its infrastructure is the one algorithms choose.