Uber Eats launches Cart Assistant, an AI system that builds grocery carts from text or image input, letting shoppers skip the manual search-and-select process.

What happened: Uber Eats introduced Cart Assistant in beta, a system that accepts a shopper's prompt or image and automatically generates a draft grocery cart with relevant items and quantities. The shopper can then review and edit before checkout. The system uses a multi-step workflow: it interprets the shopper's intent, searches store inventory, ranks products by relevance to the context, applies price and deal constraints, and selects practical package sizes.

Why it matters: Traditional grocery apps require shoppers to translate their intent into individual searches and manually pick each product—a time-consuming process. Cart Assistant shifts the workflow from intent → multiple manual searches → cart to intent → draft cart → review → checkout, eliminating the need for step-by-step product selection. This makes it faster and simpler for shoppers who start with a recipe, a list, or a vague meal plan.

What to watch: The system splits decisions between AI reasoning and deterministic logic. LLMs handle ambiguous requests (like interpreting "pasta for two" or "gluten-free lasagna"), while deterministic backend systems handle retrieval, pricing, eligibility, and cart construction. Guardrails separate grocery-shopping requests from unsupported intents and guard against prompt injection. Cart Assistant is currently in beta on Uber Eats.