Your Next Customer May Never See Your Website

Because the shopping journey may increasingly be handled not by customers themselves, but by AI agents acting on their behalf.

For years, we have been optimizing the same journey in digital commerce.

The customer searches on Google. Sees an ad. Visits the website. Examines the product. Compares prices. Adds it to a cart. Fills out a form. Pays.

From SEO to user experience, from call centers to marketplaces, a vast economy has been built on one assumption:

There is a human on the other side of the screen.

That assumption is beginning to change.

Muse, introduced by Meta in September, can now send emails, book travel, and complete tasks on the web on its user’s behalf. In July, Visa announced that live transactions had begun in Europe in which AI agents selected and purchased products on real commercial websites within limits set by cardholders. In other words, the ‘one day’ in ‘One day, an agent will shop for you’ is gradually becoming today.

What is interesting here is not that shopping is being done with AI.

For companies, the real question is:

Can customers buy from you without ever coming to you?

Email will not die. But you may not be the one reading it.

Some discussions of agentic AI suggest that email, phones, and even marketplaces could disappear.

I see the issue somewhat differently.

They do not need to disappear.

People may simply have less need to use them personally.

You do not open ten tabs to book a hotel. You tell your agent your budget, dates, and preferences.

Before you hold demo calls with ten companies to buy new software, your agent can compare their products, prices, and contract terms.

Instead of requesting quotations from a hundred suppliers, a procurement manager defines the criteria; software can carry out the first screening round on their behalf.

The need still belongs to a person.

So does the money.

The person sets the boundaries of the decision.

But a significant part of the transaction journey moves to the machine.

Data published by Cloudflare on September 30 offers an early indication of the scale. AI-agent requests on the company’s own network have risen by more than 1,700% over the past year. Cloudflare also says that more than half of the internet traffic it handled on its network this year was not human-generated. Of course, not all of that comes from AI agents; it includes traditional bots and other automation. But the title Cloudflare gave its article captures the issue well:

‘The internet now has a second audience.’

This second audience may not be your customer.

But it may be acting on your customer’s behalf.

That is a major distinction.

The new question for companies: ‘Will the agent choose us?’

Many things make sense on a website designed for people.

Photography.

Color.

A slogan.

A menu.

A promotional banner.

‘Only two left.’

An agent needs something a little different.

What is the actual price?

Is it in stock?

When will it arrive?

What is the return policy?

Are product specifications available in a structured format?

Does it meet the customer’s criteria?

Can the transaction be completed programmatically?

That is why the infrastructure companies such as Google and OpenAI have introduced in recent months should not be seen merely as technical protocols.

Google’s Universal Commerce Protocol, developed with companies including Shopify, Etsy, Wayfair, Target, and Walmart, aims to give agents and merchants’ systems a common language, from product discovery through payment and after-sales service.

OpenAI’s Agentic Commerce Protocol approach similarly relies on making merchant catalogs, inventory, and product information directly understandable to AI.

These developments remind me of the early years of the internet.

Back then, companies were asked:

‘Do you have a website?’

That was the question.

Then it changed:

‘Are you mobile-friendly?’

‘Can people find you on Google?’

Soon, another question may follow:

Can agents understand you?

Data space may replace shelf space

This change has an even more interesting implication for marketing.

Today, a brand fights to win a customer’s attention.

Tomorrow, if the customer’s agent can scan hundreds of alternatives in seconds, part of the attention economy may become an economy of fit.

An agent may not be impressed by the word ‘premium.’

But it can compare warranty periods.

It may ignore a ‘most popular’ label.

But it can examine the distribution of customer reviews, delivery performance, and total cost of ownership.

This does not eliminate the importance of brands. A trusted brand may still be a powerful signal in an agent’s decision.

But with one difference:

The story a brand tells and the reality that systems can read may have to converge much more closely.

If your product data is incomplete, your inventory information is out of date, or your prices cannot be accessed programmatically, the most beautiful website in the world may not save you.

In the past, being left off the store shelf was bad.

Tomorrow, being left out of an algorithm’s consideration set may mean the same thing.

Will marketplaces really disappear?

Probably not.

Amazon, Booking, and major local marketplaces have formidable advantages in inventory, logistics, trust, payments, and their customer bases.

But something else that matters to their economics may change:

Where the customer’s journey begins.

Today, a user opens Booking and searches for a hotel.

Tomorrow, they may tell their own agent:

‘Find me a suitable hotel in Rome.’

That may be all they need to say.

The agent may check Booking, the hotel’s own system, and other platforms.

The marketplace may still be part of the transaction, but it is no longer the first place to capture the customer’s intent.

That is no small change.

Just consider the economic value Google has created from search intent, and Amazon from product searches, over the years.

A highly valuable position is emerging in the agentic economy:

The layer that knows what customers want before they know what they will buy.

So I do not think the next platform battle will be fought solely over who has the best model.

The more critical question may be:

Who owns the customer’s agent?

In B2B, the stakes are much higher

Consumer examples are more entertaining: flights, restaurants, hotels, shopping.

In B2B, the economic impact could be much greater.

Imagine a procurement agent that knows a company’s inventory levels, cash position, supplier performance, payment terms, and foreign-exchange risk.

When a request comes in, it does more than ask for a price.

It calculates the total cost.

Assesses delivery risk.

Finds alternative suppliers.

Weighs payment terms against price.

It may even negotiate within specified limits.

Gartner has long studied this world under the heading of ‘machine customers’. In its research, CEOs anticipate that around 15–20% of company revenue could come from machine customers by 2030. In its April 2026 CEO survey, 28% of executives thought AI could threaten transaction revenues, particularly those dependent on intermediary systems, because agents could remove some layers of intermediation from purchasing, pricing, and negotiation.

We do not know whether all of these forecasts will come true.

But the direction is clear enough.

The question companies should be asking today

Not every company needs to rebuild its website ‘for agents’ yet.

But these questions now belong on the management agenda:

Can machines understand the data describing our products and services?

Are our prices, availability, and terms accessible in real time?

Can we recognize authorized software when it arrives instead of a person?

Are our ordering, payment, and after-sales processes ready for programmatic transactions?

And most importantly:

What will our competitive advantage be if our own interface no longer controls the customer’s decision journey?

In earlier phases of digital transformation, we adapted companies to technology that people used.

Web.

Mobile.

E-commerce.

Social media.

Now we are approaching a different phase.

This time, technology is not making it easier for people to reach a company.

It is coming to the company in their place.

Perhaps that is why one of the most important commercial questions in the coming period will not be ‘Are we using AI?’

It will be far more concrete:

When your customer sends a representative, will your company be able to do business with it?

Because your next customer may still be human.

But they may never see your website.