AI agents that can pick a product and start the payment are becoming practical, and businesses can set the limits. Early use is expected to be narrow.
AI software that can choose what to buy and start the payment is getting close to practical. Crypto Economy reports that AI agents are beginning to connect directly with payment and shopping systems. They work within rules set by a user or a business. An agent can prepare an order, start the payment and pause when a person needs to approve it.
Earlier AI tools mostly found products and suggested what to do. Finishing the purchase needed separate payment systems. Now agents can use outside tools more reliably, permissions can be set more precisely, and new protocols link agents with merchants and payment providers. A protocol is a shared set of rules for how systems talk to each other.
The report gives an example. A company books cloud computing capacity. Its policy limits the agent to approved providers, sets a monthly cap and requires a manager's confirmation above a certain amount. Inside those limits, the agent compares offers and prepares the purchase.
Several big names are working on different pieces. They include Google's AP2 and UCP, the Agentic Commerce Protocol from OpenAI and Stripe, and agent payment efforts from Visa and Mastercard. Together they cover connecting to merchants, recording what a user intended and approved, and identifying trusted agents.
A key point is that having a payment credential only shows the system can reach the money. Separate data is needed to show the user approved the seller, the amount and the type of purchase.
Most payment systems assume a person is at checkout. An agent may act at any hour or across several providers. The harder problem may be a purchase that does not finish cleanly. An agent might submit the same order twice, or get a payment confirmation with no matching order confirmation.
The report says adoption will depend on handling approvals, failures, revoked permissions and disputes reliably. It also notes that blockchain-based payment is one option for tiny automated purchases, such as a single API call, where normal cross-border fees could cost more than the item.
Early use is expected to be narrow: known sellers, small amounts and repeatable tasks. Business owners can still prepare now by deciding what rules they would want in place before any agent spends money. Useful questions include:
If you run a business, decide your spending limits and approval rules now, before AI tools that can buy for you reach your team.
Mastercard describes AI tools that could buy things for you, and says trust is the hard part: who gets to spend your money, and under what rules.
Read more: Mastercard outlines how AI assistants could shop and pay for youWorld models, AI that learns physics from video, are being built into home robots and driverless cars, and your own videos may help train them.
Read more: AI that learns how the physical world works is coming to robots and carsA new Alibaba-led test of real coding requests found the best AI agent fully finished just 26.4 percent of tasks, so human review still matters.
Read more: AI coding helpers fully finish only about 1 in 4 real tasks
