Until recently, "AI in e-commerce" meant a customer support chatbot or a smarter autocomplete on your search bar. This week that moved into a different category: Salesforce launched its "Shopper", "Buyer" and "Merchant" agents, integrated directly into ChatGPT, Google Search and Gemini, that search, compare and buy on a person's behalf without them ever opening a browser to visit your site.

This isn't a future promise. During last year's holiday shopping season, AI already influenced 20% of global online sales — around US$262 billion — and stores already running their own shopping agent grew sales 59% faster than the ones still sitting on the sidelines. Shopify, on top of that, launched an open protocol with Google (Universal Commerce Protocol) built specifically so AI agents can operate over a catalog without friction.

Why this isn't just another AI trend

The difference with the previous hype cycle is that this one changes who's actually browsing your site. When an AI agent buys on someone's behalf, it doesn't scroll or look at photos: it reads your catalog as data. If that data is poorly structured, incomplete, or hidden behind JavaScript that never gets to run, the agent simply doesn't find you — no matter how good your site looks to a human.

This connects to something we've been pointing at from a different angle on this blog: your site's speed matters because people leave, but now it also matters because an agent with a limited crawl budget prioritises sources that respond fast and clearly.

What your catalog needs for an agent to understand it

You don't need to "install AI" to take part in this. You need your catalog to be machine-readable, which is a lot more boring and a lot more urgent:

Real structured data (Schema.org: Product, Offer, price, availability) on every product page, not just the homepage.
A clean catalog: consistent names, no duplicate categories, stock updated in real time. An agent that recommends something out of stock burns the person's trust in a single interaction.
Content accessible without running JavaScript: if your store depends on the browser executing scripts to show the price, a fast, cheap crawl may never get to see it.
Open APIs or feeds: the easier it is for an external system to check price and availability without scraping HTML, the better your odds of being recommended.

What's not worth doing

Chasing the trend by integrating with every AI platform that shows up this week. Most of them are still defining their own standards — Shopify's protocol is months old, not years — and building on a surface that can change at any moment is betting your development budget on a single card.

What is worth doing now, regardless of whether this particular wave succeeds: clean up the catalog, structure the data, and remove the technical barriers that today keep any system — human or AI — from understanding what you sell. That never goes to waste, whatever the next trend ends up being called.

If your store depends on someone browsing it to find what they need, instead of a system being able to read it, this is a good moment to check the foundations. In our B2B and wholesale e-commerce solution we cover how we solve this technically. If you want us to take a look at your catalog, get in touch.