Key takeaway: Don’t expect a single, settled standard for agentic checkout any time soon. Visa and Mastercard are competing on their own agent-payment products while cooperating on shared identity-verification plumbing. Neither has yet spelled out how liability will be assigned when an AI agent’s purchase goes wrong.
Mastercard and Visa are each building infrastructure for AI shopping agents to authenticate and complete purchases on a cardholder’s behalf. Both want to be ahead of an expected surge in agent-initiated transactions. Most consumers, however, still won’t hand a bot the final buy click.
Mastercard rolled out a new consumer option on September 17. Cardholders can now link an AI agent to a virtual card through Alchemy’s AgentCard, the same product that has offered Visa-backed cards since June. Cardholders can cap how much the agent spends. They can also restrict which retailers it buys from, or require approval before checkout completes. Visa’s own AI shopping and payment product, Intelligent Commerce, is still being rolled out.
“This is a land grab for infrastructure standards,” Phil Bruno, chief strategy and growth officer at ACI Worldwide, told Fortune. Whichever network sets the standard for agentic commerce, he added, could keep that commerce running through its own systems for decades to come.
Infrastructure is moving faster than consumer comfort. Research commissioned by ACI Worldwide and conducted by YouGov found just 7% of surveyed U.S. and U.K. fashion shoppers would let an AI assistant buy something without their approval. More than half said they were uncomfortable letting AI purchase on their behalf at all. Shoppers valued AI most for price-drop alerts and cross-retailer comparisons, each cited by 35% of respondents. Personalized product suggestions trailed at 18%. Mastercard framed its investment as preparation for a shift it expects regardless: “AI is a transformational technology. It’s a matter of when, not if, people use it,” the company told Fortune.
The open question is what happens when an agent gets it wrong. Mastercard has built a record-keeping system called Verifiable Intent to document who authorized an agent to shop and what it was authorized to buy. The company says this could help resolve disputes. But when Fortune asked who would be responsible if an agent made an incorrect, fraudulent, or unauthorized purchase, Mastercard pointed back to that same record. It didn’t name a party accountable for losses outside it.
That liability gap sits next to a separate, broader effort. On September 10, Visa, Mastercard, and Ant International said they had begun jointly developing a Know-Your-Agent interoperability framework. The goal is to let card networks, wallets, and marketplaces recognize verified AI agents across each other’s systems, building on each company’s existing agent-verification protocols. It’s a related but distinct announcement from Mastercard’s September 17 product launch.
Why it matters: Merchants are left absorbing ambiguity while authentication and liability rules for agent-initiated purchases remain unsettled. A bot might misread a budget. An agent account might get compromised. A shopper might dispute a purchase they say the AI made in error. In each case, who’s on the hook is still unclear. Whoever’s standard becomes the default for verifying agents will also shape who controls the checkout relationship as agent-initiated transactions scale. McKinsey has projected agent-driven consumer commerce could reach $3 trillion to $5 trillion globally by 2030.
Source: Fortune; Ant International; Business Wire
