Imagine an AI research assistant comparing three locations for a new café. It has rent, foot traffic estimates, and nearby competitors. One missing fact would settle a question, but it lives in a specialist data service that charges twenty cents for the answer. The assistant requests it, pays within its owner’s budget, and folds the result into its analysis.

One question. One answer. One tiny purchase. At that size, the pricing page can cost more attention than the answer itself. Machine payments let software buy a defined service at the moment the work calls for it.

There are two kinds of transaction here, and the difference matters. When an assistant books a haircut, a person is the customer and the appointment is the goal. When an agent buys a data point, the software consumes it while doing a larger job. A person or business still funds and controls the agent, but the seller’s immediate customer is a program asking for one thing.

The twenty-cent problem

Most of the web still wants to sell a subscription. A program needing one image analysis, one weather reading, or one parcel quote is invited to create an account and choose a monthly plan. Imagine doing that at a vending machine before it will sell you a bottle of water.

The x402 protocol offers a cleaner exchange. A program asks for a resource. The server responds with HTTP 402 Payment Required and the price. A compatible client pays and retries; the server verifies payment and delivers the resource. The standard can accommodate different payment schemes and networks. What matters to the buyer is that the price and the answer arrive as part of the same request.

Coinbase documents a catalog where an agent can find and call paid endpoints; the client handles the payment challenge and retry. Its wallet instructions also show a maximum amount for a paid call. This is a working pattern for supported services, with the price check sitting where software can use it.

The rails are still expanding. In September 2026, the x402 Foundation merged a specification for Bitcoin Lightning payments. For builders of paid digital services, it adds another method they can implement alongside existing x402 payment options.

A vending machine with a better receipt

Consider a specialist business that maintains valuable information: a building supplier’s live stock list, an analyst’s narrow research feed, a company that can quote a shipment. It might already sell reports or answer requests by email. A small slice of that value could have a price and a clear delivery format. Then an agent assembling a project can buy precisely what it needs, precisely when it needs it.

The interesting unit may be tiny. A logistics agent needs one freight estimate while assembling a larger quote. A design tool needs one licensed texture. A research workflow needs one current filing extract. None calls for a monthly relationship with every seller. The hard commercial question is whether the answer is worth buying and can be delivered reliably at that scale.

Now imagine selling the same answer to a hurried stranger a thousand times a day. The description has to say what the buyer gets, how fresh it is, what it costs, and what uses are allowed. If the source was updated yesterday, say so. If the quote expires in an hour, include the clock. Payment must be verified before delivery. A failed request should explain what happened; a retry should not charge twice for the same answer. The operator funding the agent needs a budget and a record of every purchase. At twenty cents each, a bad loop can still make a painful bill.

That is the more interesting picture than an autonomous robot with a wallet. A small, useful answer is priced, bought, delivered, and accounted for inside a larger piece of work. The receipt matches the answer. The software can move on.

What this means for a business with human customers

If you run a salon, bakery, or garden studio, your first opportunity is still a human customer. Their assistant needs to understand your offer and use the booking or checkout services you already run. If you own valuable digital information or a service that software can consume by the request, machine payments open another kind of storefront. Those two opportunities can coexist without sharing the same payment setup.

Both stories reveal the same pressure on the web. Software increasingly asks for a specific thing, checks the terms, and tries to complete the action. For a data seller, that action might be a paid answer. For a salon, it is an appointment on the calendar. Being mentioned in an AI response can start the conversation; the business opportunity lives in what happens after the mention.

Look at your business through the assistant’s eyes: what can it understand, what can it offer, and where does the action stop? The free AI Visibility Audit starts with the facts on your website and helps you find the customer path worth examining next.