How MIX-STROY qualifies 100+ leads a day across five inbound channels.

The problem: 100+ product enquiries a day without a usable handoff

MIX-STROY sells construction and landscaping materials. Customers commonly begin with a photo and a short question about price, stock, size, or availability. With more than 100 inbound enquiries a day arriving through several messaging channels, the sales team had to repeat product questions, check stock manually, and re-enter each opportunity into the CRM before they could begin the real sales conversation.

What we built: one qualification layer across five channels

We set up the Bitrix24 sales funnel and connected one AI-led qualification workflow to Instagram, WhatsApp, Telegram, Avito, and MAX. The workflow identifies the product line from a photo or description, shares relevant specifications, and asks the same practical questions a good sales representative would ask before quoting.

It reads live stock levels and retail or wholesale pricing from Bitrix24 rather than inventing an answer. It can distinguish a retail request from a contractor or bulk order, capture details such as size, design, colour, quantity, source, urgency, and price objections, then create a structured CRM record for the team.

What changes for the sales team

Every qualified lead is routed to the nearest warehouse based on the customer’s city and handed to an operator with the key context already recorded. Reps spend less time turning chat fragments into CRM data and more time quoting an opportunity that is already scoped.

The live system handles more than 100 leads per day. It also cross-sells once, at the end of the conversation, when there is a relevant complementary product rather than interrupting the qualification flow.

What the demo shows

The Loom walkthrough and Bitrix24 board show the operating workflow: new enquiries move through AI qualification and into a connected-operator stage. It is evidence of the CRM and messaging handoff, not a generic chatbot demonstration.

Limits we designed around

The system is only as reliable as the product catalogue and retrieval layer behind it. Early product-reference gaps caused a small number of misidentifications, so the catalogue data was tightened. Product retrieval also becomes less reliable as the catalogue grows past roughly 100 items without stronger retrieval design.

Conversationally, fast multi-message chats can still feel unnatural. Debouncing and multi-message handling improve that experience, but they do not remove the need to monitor real conversations and refine the workflow over time.