On November 16, 2026, Commerce, the parent company of BigCommerce, introduced Shopping Agent. It is an AI assistant that a merchant builds and brands for their own storefront, and it talks shoppers through choosing a product. Commerce says it is rolling out to eligible BigCommerce merchants now, ahead of the holiday selling season, and will reach more regions and product categories over time.
This is a different thing from the AI shopping that happens inside ChatGPT or Google. Those agents send shoppers to your products from somewhere else. Shopping Agent sits on your own site, uses your catalog, and keeps the conversation and the customer with you.
What Shopping Agent does
According to the announcement, the assistant can:
- Narrow down products around what the shopper needs, likes and plans to use them for.
- Compare products and pull in the details that matter for a decision.
- Keep track of the conversation, so a follow-up question builds on the last answer.
- Match the store’s look through its own design and customization tools.
- Add items to the cart and guide the shopper toward checkout from inside the chat.
Commerce gives two examples: someone choosing a gift for a new hiker, and someone looking for a jacket for a rainy commute. Toner Buzz, a BigCommerce merchant, describes its version of the problem: customers arrive with a printer model, a cartridge number or just a question, and the goal is to turn each of those into a purchase with fewer searches.
Commerce also points to the data side. The conversations give the merchant first-party signal about what shoppers ask for and how they describe it. For a store that has only ever seen search terms and bounce rates, that is new information.
Conversational search is the part every paid plan gets
Behind the assistant is a new conversational storefront search, which Commerce says is now available on all paid BigCommerce plans. It combines keyword matching with semantic understanding, so a search written in everyday language can still find the right products. Shopping Agent uses it to fetch options while it works through a shopper’s question.
This matters even for stores that never turn the chat on. Site search on most stores still fails on plain phrases like “waterproof jacket for cycling” when the product titles say “rain shell”. If your store is on a paid plan, check whether the new search is active and run the ten phrases customers use most, taken from your search reports or support emails.
Who it is for, and what is still coming
Shopping Agent is built for B2C retail. For business buyers, Commerce is working on a separate B2B Agent, now in early access, which handles complex catalogs, negotiated prices and account rules. Commerce says AS Colour already uses it to automate purchase order workflows.
“Eligible merchants” is not defined in the announcement, and the rollout is gradual. The Shopping Agent product page is the place to check whether your store and region are included.
What we would check before turning it on
An assistant can only be as good as the product data it reads. If titles are vague, attributes are missing or sizes live only in images, the assistant will give vague answers with total confidence. That is worse than a plain search box, because shoppers trust a direct answer.
- Fill in the attributes shoppers compare on, such as size, material, compatibility and use case, as structured fields and not only in descriptions.
- Make sure stock levels and prices are current, so the assistant never recommends something you cannot ship.
- Write down the questions your support team answers every week and test each one in the assistant before it goes live.
- Decide what the assistant should not do: promise delivery dates, quote prices you have not published, or answer warranty questions you handle by hand.
- Read the first week of conversations. They show which questions your product pages leave unanswered.
The holiday timing is a reason to be careful as much as a reason to move fast. We would test on a small part of the catalog or on a staging copy first, and turn it on store-wide once the answers hold up.





