Openai held its annual development day on Monday, during which the company unveiled its plan to create applications in Chatgpt. The demo was impressive, showing how programs like Spotify and Figma can be called or discovered without leaving the Chatgpt window. While a large part of the world of technology turns to the integration of AI, the demo of Openai was the best image to date what could really look like an AI internet, with interfaces such as Chatgpt questioning information and performing commands directly.
If you look carefully, you may have noticed that there is a lot of room in this system so that money changes hands. Last week, the company launched the moment checkout, an agent purchase system which serves as a payment infrastructure for ad hoc purchases, connecting all the stores selling via Shopify, Etsy or Stripe. Now, applications provide the front infrastructure, allowing service providers to create their own interface in Chatgpt.
In short, Openai now has all the elements in place for the trade managed by AI, making Chatgpt a place where customers will buy and retailers will sell. This is a new huge sector of activity for the company – and with enormous implications for the technological industry. In this world, Openai does not only compete with Google and Anthropic, but also Amazon and Wal-Mart.
If you look at the OPENAI Application Applications Partners in the launch ad, you can see how far vision goes. Chatgpt can call you a taxi via Uber, book a trip to Expedia, call a plumber or locksmith via Thumbtack and order grocery products on Instacart, dishes prepared at Doordash or products in supermarkets at Target. Without too much additional work, Chatgpt could become a portal for most discretionary expenses of its users.
If it works, it would be much more than a simple subscription of $ 20 per month. The precise terms of the agreement are not yet clear, but like any application store, Openai is well placed to obtain part of all money spent on its platform. Chatgpt also recommends products, based on its wealth of data on its users, which makes the balance of powers between Openai and the retailers even more tip. In the words of Ben Thompson, Chatgpt becomes a super-aggregator, channeling customers to retailers and providing an entry point for ever greater sales volumes. OPENAI has many sectors of potential activity to exploit, but it is not an exaggeration to say that trade based on AI is one of the most lucrative options.
Openai is not the only company to have an eye on this perspective. The same day as the announcement of Chatgpt, Adobe published a report predicting that this year’s vacation would be dominated by AI assisted purchases, buyers turning to chatbots rather than search engines to find the best available offers. A distinct Mastercard report described the trade agent “new competitive arena” for finance. Google has already launched its own competitor protocol for the agent trade called AP2, which has a broader range but less dynamic than the Openai version.
The simplest version of purchases piloted by AI is to use chatgpt to find products instead of search: if you are looking for a canvas tennis shoe for less than $ 80, Chatgpt can find it for you as easily as Google search. But AI systems do not need to be passive. The AP2 specification includes a provision for purchases initiated by an agent, if you want an agent to buy concert tickets as soon as they are available, for example, or reserves a flight as soon as it falls below a certain price. Of course, there could also be agents on the other side of the transaction, merchant with buyers to obtain the best offer and willing to group the goods in the right circumstances. If retailers and customers are ready to take the plunge, changes could go far beyond a simple “buy” button.
Techcrunch event
San Francisco
|
October 27-29, 2025
The biggest unanswered question is whether the merchant public will be really interested. AI purchases are obviously attractive to OpenAi, and companies like Stripe and Mastercard also see many advantages, but users have not shown much interest in agent purchasing systems beyond the simple product search. But they did not have the opportunity; These systems are not even actually available, and it will take months before the average user can try a fully agent purchasing system.
When they finally, many depend on their reaction.
