In the 1980s, the giants of marketing were the big advertising agencies.

They owned the two things that mattered: creativity and media. They made the ads, and they bought the television time, the magazine pages, and the billboards. If a business wanted customers, it went through them.

Then the customer moved.

First the customer moved to search. Google took the moment when a person types what they want. Then the customer moved to social feeds. Meta took the hours people spend scrolling. These companies did not win by making better ads. They won by building a machine: find the person, show the thing, measure the result, and take the money — all in one loop. The agencies still made pretty ads, but they no longer owned the door to the customer. Slowly, they became suppliers to the platforms.

Then attention moved again, to short video. TikTok showed that even Google and Meta do not own the customer forever. Each era, the power sits with whoever owns the new door.

Now the customer is changing again. And this time, the customer may not be a person at all.

More and more, the one who searches, compares, and buys is an AI agent, working for a person. The person says: find me a good one, book it, order it, get it done. The agent does the walking.

Here is the problem for the old machine: agents do not watch ads. They do not scroll feeds. They do not feel a brand story. An agent reads data. It compares price, stock, time, fit, and risk. It asks questions. It negotiates. Then it finishes the task, or it walks away.

So the old way of marketing breaks in a quiet way. Paying for eyes does not work when there are no eyes. A beautiful website does not help if an agent cannot read it, cannot trust it, or cannot complete a purchase on it. Slogans do not help. Hidden fees get found in seconds. Vague quality claims get ignored, because the agent cannot check them.

Every disruption destroys one kind of value and creates another. So what becomes valuable when customers are agents?

First: being readable. A business must be findable and understandable by machines — clean data, clear offers, prices and availability an agent can parse, and ways for an agent to act, such as APIs and MCP endpoints. This is the new shop window. If the agent cannot read you, you do not exist.

Second: proof beats persuasion. Agents trust facts they can check. True price. True stock. True delivery time. Clear return rules. Real reviews. Certificates. A record of finishing jobs well. In the human era, marketing was often the art of seeming good. In the agent era, marketing is the discipline of being good, in a form a machine can verify.

Third: finishing the job becomes the product. The old platforms sold clicks. But a click is not a sale. When an agent is the customer, value sits in the completed task: the booking made, the order placed, the quote delivered. Businesses will pay for systems that take an agent all the way to done — and can prove it.

Fourth: trust and safety become selling points. An agent that can spend money needs rules: who may buy, how much, what needs a human’s approval, and what proof is saved afterward. Sellers who make this clear and safe will be chosen over sellers who are cheaper but unclear. An agent avoids risk the way a careful employee does.

Fifth: certainty earns the margin. Agents compare total cost instantly, so weak sellers will be pushed into a price war. Strong sellers will sell certainty instead: in stock, ships today, fixed total price, correct fit, and a promise that if something goes wrong, a human will fix it. Certainty is worth money in any era.

Sixth: someone must keep score. Agents change fast. Platforms change fast. A business that was agent-ready last year may fail this year. So there is a new job: test businesses the way agents see them, score the failures, fix them, and test again. Whoever owns the trusted test owns a piece of the new door.

And there is one more change, the one I find most hopeful. In the platform era, a single person had little power. Production was expensive, and distribution belonged to the giants. In the agent era, production is cheap — one person with real knowledge can build systems that serve many customers. What stays scarce is substance: real skill, real judgment, and a record of doing good work. Agents, of all customers, are built to find exactly that, because they compare proof instead of noise.

Big platforms will not disappear quietly. Google and Meta will try to move their toll booths into their own assistants, and sell position there too. That fight is coming. But in every shift, there is a window — a time when the new door is built but the new toll is not yet set. That window is open now.

The question is not whether agents will shop for people. They already do. The question is: when an agent comes looking, will it be able to find you, understand you, trust you — and finish the job with you?

That is the big opportunity. And it belongs, for once, to the people who build real things.