How to sell through AI assistants: what a store needs to make the answer, and the cart.
Friday evening, a weekend renovation ahead. Someone asks an AI assistant to pick a tool for removing old tiles, within budget, delivered by Saturday. A minute later they have three options and never opened a catalog. This is agentic commerce: the assistant does the searching and comparing. Here is what a store needs to be one of those three, and to take the shopper all the way to checkout.
How selling through an AI assistant works: four steps to a full cart.
The store provides facts: power, compatible accessories, weight, warranty. Without them, the assistant compares products by guesswork, or uses someone else’s data.
The assistant clarifies the task: what kind of tiles, how many square meters, what budget, when delivery is needed. Then it compares what actually fits.
The choice becomes an order: specific products and quantities in a cart, with the total visible before payment. The shopper confirms delivery and payment, either on the store’s site or inside the chat, depending on the channel and the connection.
A nice product page is not enough. You need data and actions.
Product attributes need a clear structure, and price and stock should come from the same source as your website. An outdated price in an assistant’s answer costs trust.
After the choice, there has to be a way to build a cart and reach checkout. A mention in an answer does not create an order. Delivery terms, order limits and what an agent may do are set by the store, not by the platform or the assistant.
To be clear about the main point: an open catalog does not guarantee that an assistant will pick your product. It makes it possible for the offer to be considered, compared and bought.
Where Shelfport fits.
Shelfport is the Opsalute product that connects an assistant to your store. You connect the catalog through a feed or API and decide what the agent may do. Shelfport passes the assistant’s requests to the store: find a product, build a cart, proceed to purchase.
In the beta we test this path on a specific store’s products. You can start with one category and one buying scenario.
After checkout, people still fulfil the order.
The order still has to be processed in the business system and picked and shipped. Helen prepares documents and matches products in the ERP; Hrusty helps plan shifts, assign tasks and pay for output. Each workflow can be addressed separately from the new sales channel.