Conversational Commerce
The practice of discovering, comparing, and buying products inside a conversational AI interface rather than on a traditional website or app.
What It Is
Conversational commerce is shopping that happens inside a chat. Instead of opening a retailer’s site, filtering a category page, and clicking through a cart, the buyer describes what they want in plain language and the assistant returns options, prices, and a way to pay without the conversation ever ending. The term predates the current wave; it described chat-based ordering on messaging apps years ago. What changed in 2026 is the interface. The assistant doing the talking is now a general-purpose AI model that millions of people already open every day for unrelated reasons, which means the storefront is no longer somewhere you go. It is somewhere you already are.
The clearest recent example is the arrival of shopping apps and ad formats inside ChatGPT. A payments company can build a product-search experience that pulls live prices from a catalog of more than a hundred million items, and an advertiser can buy a placement with a shop-now button that sits directly in the flow of a chat. The buyer never sees a category page. They see an answer that happens to be purchasable.
How It Actually Works
Underneath, conversational commerce is a language model wired to three things it does not natively have: a live product catalog, a checkout and payment path, and a set of rules about when to surface either one. The model handles intent. When you say you need running shoes for a wide foot under a hundred dollars, it parses that into a structured query. A retrieval layer answers the query against a merchant feed, and the results are folded back into the reply as cards or links. If the interface supports it, a payment step completes the purchase in place.
This is the same plumbing that powers agentic commerce, where an AI agent completes the purchase on your behalf with little or no confirmation. The difference is who holds the final click. Conversational commerce keeps a human in the loop deciding to buy. Agentic commerce hands that decision to the software. The two blur in practice, and a single product often slides from one to the other as trust accumulates.
Why It Matters Right Now
The push is being financed by a revenue problem. Frontier AI labs spend more to train and serve their models than subscriptions bring in, so they are converting their assistants into commercial surfaces. Advertising is one path, and conversational advertising is its own entry. Commerce is the other, and it is arguably the more durable of the two, because a percentage of a transaction is a cleaner business than an impression. When a lab is also preparing to answer to public-market shareholders, the incentive to turn the chat window into a checkout counter stops being optional.
For the buyer, the shift is mostly invisible, which is the point. The friction that used to mark the boundary between researching a purchase and making one is being removed on purpose.
How TWO Uses It
At TWO, the editorial line is not that conversational commerce is bad. It is that the operator should know which hat the assistant is wearing at any given moment. The same model that gives you a genuinely useful comparison can, in the next sentence, surface the option that pays the platform best, and nothing in the interface tells you which just happened. Scott’s rule when evaluating one of these tools is to ask a question he already knows the answer to and watch what the assistant recommends. If it steers toward a sponsored result, that is the tool’s true default, and it should be used for discovery but never for the final decision.
The operator decision this names is concrete. When you wire a conversational-commerce feature into your own product, you are choosing whether your customer’s buying intent is something you serve or something you sell. Those are different businesses, and customers can tell the difference faster than vendors expect.
What to Watch Next
Watch where the final click lives. As long as a person presses buy, the practice stays conversational commerce. The moment products start completing low-stakes purchases on standing instructions, it has become agentic commerce, and the disclosure and consent questions get sharper. The signal to track is not a feature announcement but a default setting: the first time an assistant ships with autonomous purchasing turned on rather than off.