What Is Agentic Commerce? A Plain-English Guide for Retailers

A growing number of people no longer start a purchase with a search bar. They start with an AI — asking ChatGPT, Copilot or a personal shopping assistant to find them something.

Agentic commerce is what happens when that AI doesn't just answer the question, but goes and does the buying too.

It's one of the most talked-about shifts in ecommerce right now, and also one of the most misunderstood. Here's what it actually means, stripped of the hype.

What agentic commerce actually means

"Agentic" comes from agent — software that can act on your behalf. Agentic commerce is commerce carried out by an AI agent for a shopper, rather than the shopper doing it themselves.

In a normal online purchase, a person browses a website, compares options, adds something to a basket and checks out. In agentic commerce, an AI does that legwork. You tell it what you want in plain language, and it handles discovery, comparison and — increasingly — the purchase itself, then reports back. The shopper sets the intent; the agent does the work.

That's the whole idea in one sentence: the AI becomes the shopper's buyer, and your store has to be legible to the AI, not just to a human.

A quick example

Say someone messages an AI assistant: "Find me a waterproof walking jacket under £150, in green, that ships to the UK this week."

A search engine would hand back a page of links to click through. An agent does something different. It interprets the request, looks across the products available to it, filters by price, colour, stock and delivery, narrows it to a few strong matches, and presents them — often with the ability to complete the purchase right there in the conversation, without the shopper ever visiting a website.

The shopper never saw your homepage, your hero banner or your carefully designed collection page. The agent saw your data.

How it works under the bonnet

Three things make this possible, and it's worth understanding them at a high level:

1.Structured product data

Agents can't "look" at a page the way a person does. They rely on clean, structured information about your products — titles, prices, variants, availability, specifications. Shopify's approach, for example, uses a system it calls Shopify Catalog to standardise and enrich that data so agents can read it reliably.

2. Transaction protocols

For an agent to actually buy, there needs to be an agreed way for it to place an order and pay. Shopify has introduced its Universal Commerce Protocol for this, and the other major AI and payments players — the likes of OpenAI, Google and Stripe — are each building their own standards for letting agents transact. This is the plumbing being laid right now, and it's moving quickly.

3. Checkout inside the chat

The newest piece is completing the purchase without leaving the conversation. Shopify's checkout-in-chat is already live in Microsoft Copilot, where shoppers can pay with Shop Pay, with more channels named as coming. Shopify says the structured product data it feeds into these channels converts at roughly twice the rate in AI chats — that's Shopify's own figure rather than an independent one, but it points to where things are heading.

Why it matters for retailers — and an honest reality check

It would be easy to read all that and conclude you need to drop everything. You don't. Here's the balanced view.

The direction of travel is real. Major platforms are investing heavily, and a meaningful share of product discovery is already shifting from search engines to AI assistants. Over time, "can an agent find and buy my product?" becomes a genuine commercial question.

But the honest reality today is that most of this is still early. Checkout-in-chat is live in a small number of channels, much of it is US-led, and the revenue most retailers are seeing through agentic channels right now is modest. This is a channel to prepare for, not one to bet the business on this quarter.

What it changes about running a store

Here's the useful part, and the thing that is worth acting on now: the groundwork agentic commerce rewards is groundwork worth doing regardless.

The stores that will do well as agents become buyers are the ones whose product data is clean, complete and well-structured — accurate titles, proper variants, filled-in specifications, sensible categorisation, structured data and metafields doing their job. In an agentic world, that data isn't back-office admin; it's your shopfront. The same investment also makes your store faster, easier to merchandise and better at conventional SEO. There's no version of this where getting your product data in order is wasted effort.

In other words, "AI discoverability" isn't a separate project you bolt on later. It's mostly the same discipline as building a well-structured store in the first place.

What to actually do about it (without overreacting)

A measured response looks like this:

• Get your product data in order — that's the highest-value, lowest-regret move, and it pays off across every channel, not just AI.

• Make sure your specifications, variants and structured data are complete and accurate rather than half-filled.

• Keep an eye on the agentic channels relevant to your market as they mature, and treat them as an emerging opportunity rather than an emergency.

• And resist the urge to tear up a working roadmap for a channel that, for most retailers, is still finding its feet.

If your product data is a mess — and after years of platform migrations and quick fixes, plenty of stores' data is — that's the thing worth fixing first. It's the foundation everything else, agentic or not, is built on.

Getting your product data and store architecture right is the unglamorous work that makes everything else possible — better SEO, easier merchandising, and readiness for whatever comes next. If that's a job worth doing on your store, we're happy to take a look.

→ Shopify Store Design & Development