Shoppers forgive a blunt answer. They do not forgive an invented delivery date or a product claim the catalog does not make. E-commerce assistants have to read the order, the catalog, and the policy, and they have to hand the conversation to a person when those sources do not cover the question.
What makes AI work different here
- The order is the source for where is my package. Status, address, and exceptions come from the order system. A guessed date is a support failure.
- The catalog is the source for what the product is. Materials, fit, and compatibility come from the product data you maintain, not from a general description.
- Returns and refunds have a policy. The assistant can explain the policy and start the path you allow. It should not promise an exception.
- Cost per contact only counts if the answer was right. Deflecting a chat into a wrong answer creates a second contact and a complaint.
Where we typically start
We start from the order system, the catalog, and the policies a support agent is allowed to apply. The published e-commerce study is an order-aware support chatbot measured on cost per contact.
Services we typically provide
Case studies from this industry
Frequently asked questions
Can it promise a refund?
It can explain the policy and start the flow you have approved. An exception stays with a person.
What stops it inventing a delivery date?
The date comes from the order system. If that field is empty, the assistant says it does not know and routes the chat.
Where does the 71% figure come from?
The published order-support study: an order-aware chatbot cut cost per contact by 71%. The page states that result as that study, not as a guarantee for every store.