Broadbrand
← Back to Blog
Essay · 11 min read

The end of the product page

Naledi Khumalo

On a Tuesday morning in early May, in a flat in a Cape Town suburb, a man named Thabo Mokoena bought a kettle. He did not visit a website. He did not consult a price-comparison engine. He did not, in any meaningful sense of the word, shop. He said "the kettle is broken, get one that fits in the new kitchen and costs under a thousand rand" to a thin black speaker on his counter, and an agent — a piece of software with a name and a memory — placed the order. By midday a courier had delivered a Russell Hobbs to his door. Thabo, having paid for the kettle without ever seeing a product page, was already at work.

This is what agentic commerce looks like in the spring of 2026: the human has left the building. The buying happens. The brand has no idea who chose it, why, or whether anyone will ever notice.

For two decades the dominant ritual of digital retail was a human at a screen, scrolling. A century of consumer marketing assumed that brand exposure precedes brand consideration which precedes brand purchase. The funnel, abused and refused for a generation, was nevertheless the operating metaphor of an industry. Agents collapse it. Or worse, they replace it with a different funnel that the brand cannot see, that the brand has no role in shaping, and that the brand cannot pay to climb.

The interface that no longer needs a window

When Amazon launched its product detail page in 1995, the page was the interface. The page held the photograph, the price, the review, the buy button. Every commercial decision in twenty-five years of ecommerce was a decision about the page. Should the photograph be larger? Should the price be higher up? Should the review be summarised? Should the recommended-items rail be below the fold? Marketers, designers and engineers fought over pixels.

The page is now optional. When a shopping agent buys a kettle on behalf of a human, the agent does not look at the product page. The agent reads a feed — the structured data the brand has chosen to publish, the structured data the platform has chosen to expose, the third-party signals the agent has been trained to weight — and decides. The agent then visits the page only long enough to execute the checkout transaction. The brand's seven-figure photography budget produces nothing. The brand's painstakingly written copy is irrelevant. The recommendation widget below the fold is rendered for no one.

The product page used to be where buying happened. In 2026 it is where the agent files the receipt.

We will keep calling it ecommerce. The ecommerce of agentic commerce is not the ecommerce of 2018.

What the agent optimises for. And what it does not.

A shopping agent is not a search engine with a personality. It is a constrained optimiser. It has a goal (the user's instruction), a set of constraints (price, delivery, return policy, retailer relationships, the user's stated preferences), and a scoring function it has been trained to apply to candidate purchases. The scoring function does not weight brand love. It does not weight nostalgia. It does not weight the photograph. The fields the agent reads are the fields the agent trusts.

This is mostly bad news for incumbents. The brands that won the last twenty years won by becoming the first thing a human thought of when they thought of a kettle. They invested in a recall asset — a memory imprint in millions of human heads — and that asset compounded. An agent has no head. It has a token window. The tokens in that window contain price, fit, reviews, delivery time, refund policy, the user's stated preference, and any structured signal the brand has been able to inject. Decades of mind-share have no surface to land on.

Agent of the customer vs agent of the platform

There is no such thing as a neutral agent. Every agent is the agent of someone, and the someone is the entity that controls the scoring function. When Amazon's agent buys a kettle, it is buying as the agent of Amazon's economics. When Apple's agent eventually buys a kettle, it will be buying as the agent of Apple's economics. When a genuinely third-party agent buys a kettle, it is buying as the agent of the consumer. Genuinely third-party agents are rare.

The battle of the next thirty-six months is over which of these wins the right to be the default. The platform players will give their agents away. The third-party players will charge a subscription. The consumer will choose, mostly, by which interface is in front of them when they need a kettle.

For brands, the strategic implication is brutal. Every dollar you spend on the agent of Amazon is a dollar that strengthens Amazon's negotiating position the next time it sets terms. Every dollar you spend on the agent of the consumer is a dollar that strengthens a third party that has no marketplace lock-in. The structural pull is toward whichever agent has the most data on the consumer's behavioural pattern. Which today is a small handful of platforms.

The retail media inversion

Retail media — the eighty-billion-dollar advertising line that retailers built on top of their shopper data — was supposed to be the safe harbour. The retailer knows the basket. The retailer can sell access to the basket. The agentic shift inverts this. When an agent is doing the buying, the click-through rate of a sponsored product card is no longer the relevant metric. The agent does not click. The agent reads structured data. The retailer is no longer monetising attention; the retailer is monetising structured-data placement inside a feed.

This is not a death sentence. It is a redesign. Retail media networks built around impression and click metrics will need a parallel layer built around agent-readable promotions, agent-readable rank boosts, and agent-readable pack-deal logic. The retailers that figure this out first will sell agentic placement at premium prices for the same reason paid search rose to dominance: the platform owns the surface where the decision happens, and the brand will pay for placement on that surface.

What brands must do now

The temptation, on reading this, is to do nothing. The shopper at the screen is still doing the shopping for most categories. The numbers are still mostly human. The agentic share of category-level revenue is in the low single digits. There is time.

There is less time than there appears. The structural rewrite is happening upstream of the visible numbers. The product feed your brand submits to the agentic surfaces today is the product feed that will be ranked, weighted and trusted three years from now. The reviews you accumulate this quarter are the reviews the agent will weight next quarter. The structured-data hygiene you adopt now is the structured-data hygiene that determines whether you exist in the agent's window in 2028.

Three rules for the agentic commerce era

The brand that wins the agent of 2028 is the brand that wrote the feed of 2026.

Rule one. Stop optimising the page. Optimise the feed. The page is the receipt. The feed is the offer.

Rule two. Treat the agent the way you used to treat the buyer at a category-buying meeting. Send it specs, not stories. Give it the data the buyer would have asked for in writing — never the data you put in a deck. Anything that lives only in a slide is invisible.

Rule three. Find out who the agent is the agent of. Plan accordingly. If the agent is the agent of a marketplace, expect the terms to harden every renewal. If the agent is the agent of the consumer, expect the rules to be opaque. Either way, the spend that historically went into brand-image advertising on platforms that no longer drive purchase is the spend that ought to be redirected into structured-data infrastructure, into agent-readable promotion design, and into the operational discipline of keeping the brand's feed clean.

The shopper has left the building. Some of them will come back, for some categories. But for the kettles, the toothpaste, the printer ink, the dog food — the categories where the decision is too small to be worth the human's attention — the human is gone for good. The agent is doing the shopping. The brand that talks to the agent the way it used to talk to the shopper will not be in the basket. The brand that retools is the brand that survives the next three years intact.


Written by Naledi Khumalo ·
Subscribe →
Keep reading
Note

The AI programmatic audit is coming

May 2026 · 7 min
Opinion

What comes after the CRM

May 2026 · 9 min
Case

The new rules of the lookbook

May 2026 · 8 min