PURCHASE, N.Y.—Artificial intelligence is rapidly moving from assisting consumers with shopping to actually making purchases on their behalf, a shift that payments companies say will reshape how commerce works—and how trust is verified when machines spend real money.
In a new commentary outlining that shift, Pablo Fourez, chief digital officer at Mastercard, said AI systems are increasingly evolving into “agents” capable of planning purchases, evaluating options and executing transactions autonomously. As those systems begin acting independently for consumers and businesses, he said, the key challenge becomes verifying that the AI followed a user’s instructions exactly as intended.
“This is a new payments paradigm,” Fourez said. “And in this paradigm, trust becomes the product.”
Traditionally, digital commerce has relied on clear signals of consumer intent—a tap of a card, a click of a “buy” button, or another direct action that confirms a purchase decision. In an emerging “agentic commerce” model, however, consumers may instead delegate tasks to software agents that monitor prices, reorder goods, or book travel within certain parameters and budgets.
That delegation introduces new questions for payments networks, merchants and financial institutions. Issuers must determine whether a transaction initiated by an agent is legitimate or fraudulent, merchants must confirm that the system is authorized to act for the consumer, and consumers themselves need confidence that automated systems are carrying out instructions correctly and privately.
To address those concerns, Mastercard said it is working with Google on a framework called Verifiable Intent, a standards-based approach designed to create a tamper-resistant record of what a user authorized when an AI agent conducts a transaction. The system is intended to provide cryptographic proof linking a consumer’s identity, instructions and the resulting purchase, creating a shared record that merchants, issuers and payment networks can reference if questions arise.
The company said the framework will be integrated into its Agent Pay technology in the coming months and aligned with Google’s emerging agent payments protocols. The goal is to create a consistent trust layer as autonomous agents begin to participate directly in commerce across different platforms and devices.
Fourez said such mechanisms will be essential as AI systems begin executing transactions at greater speed and scale, including scenarios beyond typical consumer purchases. Machine-to-machine payments, automated procurement systems and cross-border transactions managed by software agents could all rely on similar verification frameworks.
“In an agentic world where actions are delegated, intent must be explicit and verifiable,” Fourez said.
The approach is also designed to address privacy concerns by allowing only the minimum amount of transaction information to be shared among parties involved in a payment. According to Mastercard, the system uses selective disclosure techniques so transaction details can be verified for fraud or dispute resolution without exposing unnecessary personal data.
Industry participants—from payment processors to technology providers—are beginning to explore standards for what could become a foundational layer of the next phase of digital commerce. As AI agents increasingly participate in transactions, the payments industry faces a central question: how to maintain accountability and consumer protection when the buyer may be software rather than a person, analysts have stated.