Shipt and Instacart each introduced an AI grocery assistant on Sept. 9, turning natural-language requests into ready-to-edit shopping carts. The parallel launches show that retail AI is moving beyond recommendations and into the transaction itself.
Ask Shipt and Instacart’s Clementine can interpret meal plans, budgets, recipes and everyday household needs. Instead of asking shoppers to search for every item, the systems assemble a cart that customers can review before checkout. That changes the role of the interface—and the commercial value of being selected by it.
What the new assistants can do
Shipt said Ask Shipt can build a cart from a prompt such as a tailgate for 25 people, suggest a family meal within a stated budget and identify ingredients from an uploaded food image. Customers can swap products, edit quantities and refine the request before placing the order.
The feature is available in the Shipt app and on Shipt.com across a marketplace of more than 100 retail partners. Shipt’s help documentation says the assistant also accepts recipes, lists and dietary preferences. The company has separately connected its shopping workflow to ChatGPT and Claude, although checkout still occurs through Shipt.
Instacart launched Clementine the same day. According to the company’s announcement, the assistant can turn a conversation, recipe or grocery list into a cart and respond to requests involving budgets, nutrition preferences and repeat purchases. TechCrunch reported that the release follows similar moves from Uber Eats and DoorDash.
The timing suggests that conversational cart-building is becoming a category feature rather than a durable differentiator on its own. Shipt’s photo input and retailer network may shape its experience; Instacart’s product data, retailer relationships and order history may shape Clementine. The important competition is shifting toward the quality of the underlying decision system.
Analysis: The assistant becomes a merchandising layer
Traditional ecommerce search gives shoppers a list of results and leaves the comparison work visible. An AI assistant compresses that process. It interprets intent, decides which products satisfy it and presents a proposed basket. Every one of those steps can affect what gets purchased.
That makes the assistant a new merchandising layer. Placement may depend on availability, price, dietary attributes, retailer promotions, previous behavior and the model’s interpretation of a vague request. Brands can no longer assume that a high search ranking guarantees consideration when the interface may choose one item before the customer sees alternatives.
The business opportunity is substantial because grocery shopping is repetitive and high frequency. If an assistant reliably converts a weekly plan into an acceptable cart, it can reduce friction and encourage loyalty to the platform. It can also increase average order value by filling in missing ingredients or suggesting products that complete an event or meal.
That convenience can become habit quickly when the assistant remembers preferences and reduces repeated planning work.
The risk is hidden error. A plausible cart can contain the wrong size, an unsuitable substitute, an overlooked dietary restriction or a promotion that makes the recommendation look more neutral than it is. The more confidently the assistant acts, the more important review controls and explanation become.
What retailers and brands should measure
Retailers should measure more than assistant engagement. Completion rate, edit rate, substitution rate, abandoned carts, repeat use and customer-service contacts will show whether the experience actually reduces work. They should also monitor whether recommendations remain accurate when inventory, prices and promotions change.
Brands need to improve the structured information that shopping systems use: product names, package sizes, ingredients, nutrition attributes, imagery, availability and clear category context. They should also ask retail partners how sponsored placement, private-label products and promotional rules affect AI-generated carts.
Transparency will become a competitive feature. Shoppers should be able to see why an item was selected, what assumptions shaped the basket and which recommendations were paid or promoted. The platforms that make correction easy will earn more trust than those that treat the generated cart as a finished answer.
Ask Shipt and Clementine are early versions of a larger shift. Product discovery is becoming an orchestration problem: understand the household’s goal, assemble the products and move the customer to checkout. For commerce leaders, the question is no longer whether AI will influence the purchase. It is how much of the purchase journey the assistant will control.