Machine Customers: How to Sell to AI Agents in 2027

For twenty years, every decision about your website, your pricing page and your checkout assumed the same thing: a human was on the other end, reading, comparing, deciding. That assumption is starting to crack. The next customer to arrive on your site may not have eyes, may not scroll, and may not care how your homepage makes them feel. It wants a price it can trust, terms it can parse, and a way to complete the purchase without a human in the loop.

Gartner has a name for this buyer: the machine customer, or "custobot." It is a non-human economic actor that can negotiate and buy goods and services on its own. The firm estimates that by 2028 there will be around 15 billion connected products with the potential to behave as customers, and that CEOs expect 15 to 20% of their revenue to come from machine customers by 2030. You can argue with the exact numbers. The direction is hard to argue with, because the agentic browsers and shopping assistants that make it possible are already in your customers' hands.

This is not a prediction post. It is a practical look at what a machine customer is, why it is not only an ecommerce problem, and what you can actually do to make your business sellable to one.

What a machine customer actually is

Strip away the futurism and a machine customer is software that buys. Gartner maps it in three phases, and it helps to know which one you are planning for.

• Bound. The machine follows rules a human set. A printer that reorders its own toner, a thermostat on an energy plan. This exists now and is boring in the good way.
• Adaptable. The machine chooses from options within a goal you gave it. "Keep the office stocked under this budget", "rebook this flight if it is delayed." This is the phase arriving now, and it is where agentic browsers and assistants live.
• Autonomous. The machine sets its own sub-goals and transacts across many suppliers with little oversight. Further out, and the one the headlines love.

Most businesses should plan for the adaptable phase, because that is the one knocking on the door. An agent acting for a real person, or for a procurement system, lands on your site with a specific job and a low tolerance for friction.

It is not just an ecommerce problem

The easy mistake is to file this under "online shops." Machine customers show up hardest in dull, repeatable B2B spend: reordering supplies, renewing software, booking logistics, comparing quotes. A procurement agent does not read your brochure. It queries your catalogue, checks stock and terms, and shortlists you, or does not, in milliseconds.

That is the same shift we wrote about in how to get your business shortlisted by AI buyers. If a human never sees your sales page because an agent filtered you out first, your beautiful copy never gets a vote. Services businesses are exposed too: an agent trying to book a slot, get a quote or start an onboarding hits the same walls a shopper does, just with less patience and no willingness to phone you.

Why your business is unsellable to an agent right now

Most sites are built to persuade a human and quietly hostile to a machine. The failure modes are mundane and fixable.

• Prices locked in images or behind a form. An agent cannot read a price baked into a JPG or gated behind "contact us." If the number is not in text, you are invisible to the comparison.
• Checkout that needs a human hand. Hover menus, multi-step flows that depend on visual cues, a CAPTCHA that stops the agent cold. We covered the mechanics in is your website ready for AI browsers and agents.
• No machine-readable facts. Stock, delivery time, return policy, terms. A human infers these from context; an agent needs them stated plainly, ideally as structured data.
• No path to transact without a browser. The most reliable agents do not click buttons at all; they call an interface. If you expose nothing, you force the fragile route.

How to make your business sellable to an agent

You do not need to rebuild everything. The work layers from cheap hygiene up to real engineering, and most businesses should stop well short of the top rung for now.

1. Put the facts in text. Prices, availability, lead times, terms, written plainly and machine-readable. This is the single highest-return move and it also helps you rank in AI search and agentic commerce.
2. Make the human flow robust. If an agent is going to drive your normal checkout, it has to survive without visual guesswork. Predictable forms, no CAPTCHA wall on the buying path, clear confirmation states.
3. Expose actions through an interface. For anything repeatable (a quote, a reorder, a booking), a defined interface beats screen-clicking every time. This is the idea behind WebMCP and the agent-ready website: give the agent a real door instead of making it climb through the window.
4. Decide your payment story. Agent-initiated payments are maturing fast. If machine orders are coming, you need a view on how they pay and how you stay in control, which we unpack in AI agents, payments and checkout.

What to do this quarter

You do not need a machine-customer strategy offsite. You need three small moves. First, tag the agentic traffic you already have so the conversation starts from data, not hunch. Second, run the ten-minute test: point an agentic browser at your own site and tell it to buy, book or quote, then fix whatever stops it. Third, pick the one repeatable transaction that matters most to your revenue and ask whether it deserves a proper interface rather than a hope that the agent clicks correctly.

The businesses that win the machine-customer shift will not be the ones with the flashiest AI announcements. They will be the ones whose prices, stock and checkout are boringly easy for a machine to read and complete, the same way the early web rewarded whoever simply had a working online order form. If you are weighing how much of your site or stack to adapt, and how much to leave for later, tell us how your business works and we will help you separate the part worth doing now from the part worth watching.