Retail Brand — AI Shopping Assistant

Most storefront chat is either a rules-based decision tree that frustrates people or a support inbox nobody staffs at 11pm. Neither helps the customer who knows what room they are furnishing but not which product fits it.

I built a conversational assistant that sits on the storefront and answers with knowledge of the actual catalogue — size, material, room suitability, care — and moves the conversation toward a specific product rather than a category page. Brand voice, tone and product framing are configuration, not hard-coded, which is what let a second brand deploy on the same codebase without a fork.

Two rules are enforced at the model boundary: never mention a competitor, and never link outside the store. Beyond correctness, that is a commercial requirement — an assistant that helpfully sends a customer somewhere else is worse than no assistant at all.