The Problem
An e-commerce retailer had a catalog rich with product attributes — effects, uses, chemical profiles, and ideal usage times — but shoppers rarely know product names. They know how they want to feel and what they want help with.
Customers also asked open-ended consultation questions that a simple product filter could never answer, leaving a gap between browsing and expert guidance.
Our Solution
We built a dual-mode retrieval-augmented chatbot that both recommends products and acts as a knowledgeable consultant.
Recommendation mode
The shopper describes the effects and uses they want, plus product type and timing. The system semantically searches the catalog, then an LLM ranks the best matches on relevance and returns real products from the database — never invented ones.
Consultation mode
A conversational consultant answers open-ended product and usage questions with multi-turn memory, grounded in a dedicated knowledge base, and politely declines off-topic queries.
Hallucination-proof by design
The AI never emits product descriptions freely — it returns validated product IDs that are looked up from the source-of-truth database, so every recommendation is a genuine catalog item. Metadata pre-filtering ensures results always match the shopper's stated preferences.
Outcome
Shoppers describe what they want in plain language and receive ranked, real product recommendations plus expert consultation — turning a static catalog into a guided, conversational buying experience with zero hallucinated products.
Tools & Technologies
Build something similar
Start with a free requirements assessment — we'll tell you exactly what's achievable.
Start Free Assessment