Your next customer may not be human

For most of marketing history, the path to a sale began with a person.

Someone noticed an advert, opened a review or heard a recommendation from a friend. However long the journey, people encountered the options and decided what deserved their attention.

Now people are asking AI systems which laptop to buy, where to stay or what software fits their business. The payment may still be made by a person, but the preceding work can happen elsewhere. A system searches, compares and excludes. By the time the customer sees a shortlist, dozens of brands may have disappeared.

Most tools doing this today are AI assistants. They make recommendations while the person retains control. An autonomous agent can go further, carrying a task across systems and acting within permissions set by its user. Both can influence which brands are considered, even when no purchase is delegated.

Brands may soon have to earn their place twice, first with the system filtering the market and then with the person making the decision.

The invisible shortlist

Search engines changed how brands were discovered. Social platforms changed how they earned attention. AI agents change the point at which judgement enters the journey.

A traditional search presents choices. An AI system can interpret a request and return a conclusion. Someone might ask for accounting software suitable for a small Portuguese company or a quiet hotel in Paris. The system translates those preferences into criteria and decides which options are worth presenting.

A largely invisible stage is taking shape in the buying journey, where brands can be assessed before they have an audience.

McKinsey’s research into European consumers describes a market in which AI already influences decisions more readily than it executes purchases. Public discussion tends to focus on bots placing orders autonomously. The more meaningful shift may happen earlier, while AI is deciding which brands make it onto a human shortlist.

Built for attention, assessed through evidence

Marketers have spent years learning how to gain attention and build audiences. Those skills were developed around human perception. An AI system does not admire art direction, feel urgency or remember a jingle from childhood. Its assessment depends on the information it can access and interpret.

Product data, availability, price, independent reviews, return conditions and consistency across sources can influence whether a brand survives the filtering process. Stripe’s early work in agentic commerce has found that reliable product information is central to appearing across agent-led shopping environments. Even inventory needs to be accurate in real time.

This gives operational clarity a role in brand visibility. Vague product pages create interpretation work. Conflicting specifications weaken confidence. Unsupported claims become difficult to distinguish from thousands of similar claims.

The polished campaign still has value because people continue to form preferences, identities and emotional attachments. Yet the systems helping them choose may reward a different layer of the brand, built from what can be verified.

Whose interests does the system serve?

Marketers will naturally identify the signals that improve visibility and optimise around them. The risk appears when satisfying the system gradually takes priority over the person it represents.

An agent is meant to represent the interests of its user. It may know that the customer values durability over novelty, avoids subscriptions, prefers local suppliers or will pay more for a generous returns policy. The better these systems become, the less useful generic persuasion may be. A brand will have to fit the actual conditions of the request.

This could make some parts of marketing more honest. When attention is less available as a shortcut, evidence becomes harder to avoid. Reputation needs support from product quality, service records, transparent policies and credible third-party accounts.

It also raises uncomfortable questions. Who decides which sources an agent trusts? How are commercial relationships disclosed? Can a brand pay for preference inside a recommendation that appears objective? The interfaces may feel conversational, but the infrastructure beneath them will carry commercial incentives of its own.

A second customer enters the room

Human desire will remain at the centre of commerce. People will still want beautiful things, familiar names, reassurance and stories that help them understand a choice. Those desires may increasingly be interpreted by a machine before a brand gets the chance to respond.

Marketing has always adapted to intermediaries: retailers, media owners, search algorithms, marketplaces and creators. AI agents belong to that lineage, with one important difference. They can act as researcher, adviser, gatekeeper and buyer at the same time.

The brands that remain visible will need more than memorable communication. They will need to be legible, credible and useful to systems that decide what a person should see.

The customer is still there, although reaching them may increasingly require passing through a system’s evaluation.