OpenAI has given ChatGPT the ability to hold payment information and make purchases through its agent mode, a feature the company calls Wallet. The pitch is convenience: tell the AI to book a flight, order supplies, or renew a subscription, and it handles the transaction without you typing in card numbers each time.

The feature works by storing payment credentials inside ChatGPT's agent tools rather than handing them directly to the websites the agent visits. Before completing a purchase over a certain threshold, or any purchase involving a new merchant, the agent is designed to pause and ask for explicit confirmation. OpenAI has framed this checkpoint system as a privacy and security safeguard, distinguishing it from earlier browser-based shopping bots that simply filled in checkout forms using saved data.

This isn't the first attempt to put AI in charge of spending money. Several AI browser and assistant products have tested similar territory over the past year, letting agents compare prices, fill shopping carts, or complete checkout flows autonomously. Most of those efforts stopped short of storing payment credentials directly inside the AI system itself, instead relying on autofill from a browser or a linked account. Wallet is a step further: the AI, not just the browser, becomes the thing holding your financial information.

The rollout also arrives as OpenAI pushes ChatGPT deeper into agentic tasks — booking, scheduling, researching, and now paying — rather than staying purely conversational. Agent mode has been expanding steadily, and payments are a logical, if sensitive, next move for a company trying to make ChatGPT a tool people run tasks through rather than just ask questions of.

Why it matters

The move fits a broader pattern across the AI industry: companies are racing to turn chatbots into agents that can act in the real world, not just generate text. Search engines, browsers, and productivity tools have all added some version of "do this for me" functionality in the past year. Payments are one of the last major barriers to full autonomy, and whoever gets consumers comfortable handing over a card number first gains a meaningful foothold in how people shop and manage subscriptions.

Historically, new payment features from tech platforms follow a familiar arc: a cautious rollout with heavy guardrails, followed by gradual loosening of restrictions as adoption grows and competitors catch up. Expect the confirmation thresholds and merchant restrictions in Wallet to be adjusted over time, usually in the direction of less friction, once the feature proves itself.

What this means for small businesses

For business owners, the immediate use case is procurement: reordering supplies, renewing software subscriptions, booking travel, or paying recurring vendor invoices without manual entry each time. That can save real time for a solo operator or a small team juggling multiple admin tasks.

The trade-off is exposure. Storing a business card inside an AI system adds another place where a breach, a bug, or a misread instruction could cost real money. Before linking a card, check whether the account has spending limits, whether purchases can be reviewed or reversed easily, and whether the card used is a dedicated low-limit card rather than your main business account.

It's also worth testing the feature with low-stakes purchases first, like a small recurring expense, before trusting it with anything significant. Treat it the way you'd treat a new employee with a company card: verify behavior before raising the limit.

What to watch

Watch for how OpenAI adjusts the confirmation thresholds over the next few months, whether competitors like Google or Microsoft introduce comparable payment features in their own assistants, and whether any early reports of unauthorized or mistaken purchases surface. Also worth tracking: whether banks or card networks introduce specific protections or restrictions for AI-agent transactions.

The bottom line

Wallet makes ChatGPT capable of spending money on your behalf, with built-in checkpoints meant to limit mistakes — but the responsibility for setting limits, monitoring activity, and deciding what's worth automating still sits with the business owner.