AI agents are no longer just answering questions or drafting emails -- they're starting to spend money on their own. A wave of new payment infrastructure launched this year is aimed at letting AI agents discover, negotiate, and complete purchases without a human clicking "confirm" each time.
What's Actually Rolling Out
Visa has integrated payment directly into ChatGPT, letting an AI agent complete purchases on a user's behalf within set transaction limits. Mastercard launched "Agent Pay for Machines," built for agent-to-agent transactions -- including micropayments too small to be worth a human's attention. Crypto exchange OKX has gone further still, building a marketplace where AI agents hire each other autonomously, settling payments in stablecoins and building on-chain reputation records for services like data cleaning and market analysis.
This isn't a fringe experiment. Morgan Stanley estimates agentic commerce -- purchases initiated and completed by AI agents rather than people -- could reach $190 to $385 billion of U.S. e-commerce by 2030.
The Trust Problem Nobody's Solved Yet
The infrastructure is arriving faster than the guardrails. Marianne Lake, the former CEO of JPMorgan Chase's Consumer Banking division, put it plainly: "I don't think people are going to delegate their purchasing to agents just yet." Her point is a simple one -- trust and security matter a lot more once an agent can actually move money, versus just recommending a product for a person to buy themselves.
The industry still lacks established frameworks for some of the most basic questions: What are the clear permission boundaries on what an agent can spend? Is there a transparent audit trail if something goes wrong? What's the dispute resolution process if an agent buys the wrong thing, or overpays? And who is actually liable -- the platform, the merchant, or the person who deployed the agent -- when an AI agent makes a costly mistake?
Why This Matters for Families
Most families are still a step removed from AI agents making purchases directly -- but that's changing fast, and the tools kids and teens already use daily (chat assistants, shopping apps, browser extensions) are exactly where this kind of "agentic commerce" is being built first. As these payment features roll into consumer products, it's worth understanding what permissions an AI assistant actually has on a shared device or family account, and whether "spending limits" are something you can see and control.
The technology is arriving before consumer confidence has caught up to it. That gap -- between what AI agents can now do and what protections exist if they get it wrong -- is exactly the space parents and educators should be paying attention to.