Guide
How do AI shopping assistants work?
Updated · 11xsales.ai
Short answer
An AI shopping assistant combines your store data with a language model. When a shopper asks something, it searches your catalog, policies and pages for the relevant facts, and the model writes a reply using only those facts, then shows matching products and lets the shopper add one to the cart. Guardrails check prices and keep it from inventing products or policies.
The steps behind one reply
- Read: the store's products, variants, stock, policies and pages are indexed so they can be searched by meaning, not just keywords.
- Understand: the model reads the shopper's message, the conversation so far and the page they're on.
- Look up: it searches the catalog or knowledge for what the question needs, for example in-stock products under a price.
- Answer: it replies in the shopper's language from what it found, recommending a few products with a reason each.
- Act: product cards let the shopper pick a size and add to cart; order questions use the order number and email.
- Hand off: when a person must act, the conversation goes to your team with its full history.
Why grounding matters
Language models can sound confident about things that aren't true. A shopping assistant should look things up before answering and check facts like prices against the catalog. 11xsales.ai checks the product prices it writes against your catalog and corrects a wrong one before the reply is sent.
What the merchant controls
You choose what the agent knows (catalog, pages, documents, corrections), how it sounds, when it brings in your team, and where it appears on your site. Chats from your own preview are free, so you can test before shoppers see it.