AI shopping advisor in an on-site chat

The problem

An online store selling climate equipment: it is hard for a visitor to pick the right model for their room, and a classic filter or quiz puts people off. The manager receives «cold» leads with no data about the room and wastes time asking the basics. What was needed wasn't a robot salesman or a questionnaire, but a polite assistant that helps the customer choose — and, along the way, prepares a qualified deal for the manager.

Before

• Choosing a product was on the customer: dozens of models, specs, power sizing for the room area.
• Leads arrived with no context: the manager re-established the area, conditions and budget from scratch.
• Cross-sells relied on the manager's memory and often simply didn't happen.

What we built

• We built an AI advisor as an on-site chat: it holds a natural conversation rather than firing off a list of questions, and casually collects around 20 room parameters — area, ceiling height, sun-facing side, windows, floor, budget, need for quiet, Wi-Fi, installation and so on.
• The AI does not invent products: the catalog is pulled from the store's live product feed, and the backend itself computes the required capacity with corrections (sun, panoramic windows, top floor, people and equipment in the room) and ranks real items.
• The selected models are shown right in the chat as cards — with the real price and a link to the product.
• When the customer leaves a phone number, the assistant creates a contact and a deal in Bitrix24 with a structured summary for the manager — room parameters, needs, a capacity recommendation.
• The deal also carries a follow-up plan hidden from the customer (for example, offering a heater in the off-season) — the manager sees it, the customer does not.

Results

• The visitor gets a selection from the real catalog rather than a rough suggestion: products, prices and links are live, from the store's feed.
• The manager receives an already-qualified deal with the full room context — no need to re-establish the basics.
• Cross-sell opportunities are captured automatically in every deal instead of being lost.
• The dialogue feels like help from a consultant, not a CRM questionnaire — the internal goals stay hidden from the customer.