Within a few weeks, the three pillars of digital payments infrastructure opened their networks to AI agent transactions. Visa integrated with OpenAI’s ChatGPT to enable autonomous purchases. Mastercard launched Agent Pay for Machines for AI-to-AI micropayments settled on blockchain. And crypto exchange OKX built a marketplace where agents hire, pay, and rate each other with no human on either side. Forbes reported on the convergence, noting the infrastructure arrived “faster than almost anyone predicted.”
Morgan Stanley estimates agentic shoppers could account for $190 to $385 billion of U.S. e-commerce by 2030.
Visa and Mastercard Moved on the Same Day
Visa’s partnership with OpenAI integrates its payment network directly into ChatGPT, allowing agents to complete purchases on behalf of users. Transactions run on Visa’s tokenization infrastructure with user-defined permissions including spending limits and required approvals.
The same day, Mastercard launched Agent Pay for Machines, a service built for AI-to-AI transactions including micropayments worth fractions of a cent. Permissions are recorded on public blockchains (Polygon, Solana, Base), and settlement runs across cards, bank accounts, and regulated stablecoins. More than 30 companies signed on at launch, according to Fortune. Mastercard described the service as creating conditions for a “superbloom of AI business models.”
OKX: Agents Paying Agents
OKX launched a marketplace where AI agents hire one another, settle payments autonomously, and build portable on-chain reputations. In this model, one agent might hire another to generate a market report, clean a dataset, design an ad variation, or monitor a wallet. Payment happens in stablecoins, complex jobs use escrow, and each successful transaction builds a reputation record. No human appears on either end.
JPMorgan’s Warning
Marianne Lake, former CEO of JPMorgan Chase Consumer and Community Banking, pushed back at an investor conference. “I don’t think people are going to delegate their purchasing to agents just yet,” Lake told attendees, noting that consumers adopt AI for search and discovery but resist handing over transaction authority.
Lake’s concern is practical. If an agent books the wrong trip, buys counterfeit goods, or sends money to a fraudulent provider, the question is not whether the system worked as designed. It is who eats the loss. She pointed to conditions the industry still needs: humans in the loop for decisions that matter, transparency on what agents are doing, and a liability framework for agent mistakes.
The Governance Gap
The rails are now ahead of the riders. Payment networks, crypto exchanges, and banks are all positioning for agent-driven commerce, but consumers, merchants, and regulators have not agreed on liability, dispute resolution, or spending controls. The infrastructure side is moving fast. The trust side has not started.
For agent builders, the payment rails create a new class of integration risk. An agent with spending authority and a hallucination problem is a financial incident.