AIBP

The company had amoCRM before us, but nobody reviewed the conversations: the manager listened selectively, the real reasons for refusals stayed inside the calls, and inquiries got lost when handed from one person to another. In April 2026 we started with a single area — reviewing conversations — and since May it has grown into a platform the whole team works on.

AI listens to every conversation and writes the review straight into the deal card, puts together a dossier on the client before the call, and after the call suggests the next step and sets the task. Each employee has their own workplace with a plan for the day: one qualifies incoming inquiries, another works deals stuck for 45 days, a third warms up clients who went cold — and there a letter only goes out once a person has read it and pressed the button.

A commercial proposal is put together by voice in a minute instead of half an hour in an editor. Separately, you can see what each employee did during the day — the same kind of review, only of the work rather than of the clients. The platform is refined every week: it now holds four workplaces and covers the client's whole path from the first inquiry to installation.

• Every conversation reviewed into the deal card
• A dossier on the client before the call
• The next step and a task after the call
• A workplace with a daily plan for everyone
• Qualification of incoming inquiries
• Working deals stuck for 45 days
• Warming up cold clients, letters confirmed by hand
• A commercial proposal by voice

The company sells products under its own brands on Ozon — Quorumdent, Royal Tools, Barocco, Lamucha. Other sellers kept latching onto its product cards: they used the registered trademark and the brand’s content and lured buyers away with lower prices. On one toothpaste, seven Chinese shops sat at 777 ₽ against the brand’s 1,673 ₽. A dedicated employee spent four hours a day finding violators by hand and writing complaints.

We built a robot that checks the whole catalogue in one pass: it takes the product list from the seller account, opens every card the way a buyer sees it and finds every third-party offer — with its SKU, legal entity, country, price and delivery date. The robot sees that thirteen storefronts belong to eleven companies, and it tells the brand’s own accounts apart by a list rather than by the marketplace’s unreliable badge.

Then the robot goes through the whole complaint path in Ozon support on its own — from the Quality Control section to the trademark certificate — and files complaints one after another in a single ticket. If the marketplace asks it to wait, it pauses by itself. Every complaint goes into a log: when, against which shop and which legal entity. On the very first day, all seven third-party shops were removed from the hardest-hit card.

• The whole catalogue checked in one pass
• Third-party sellers on every card
• Violator’s legal entity, country, price and delivery date
• Storefronts traced to real companies
• Complaints to Ozon support without a human
• A batch of complaints in one ticket
• Pauses when Ozon asks to wait
• A log of every complaint

Review from the company owner

The checkout runs on a phone, tablet or laptop: the cashier scans a QR code or enters a promo word, and the discount and earnings are calculated automatically. A payment cannot be deleted, and a refund requires a reason. Cashiers log in with a PIN and each has shifts, a log and end-of-shift reconciliation.

Partners and ambassadors see their balance in a personal account and get a message about every accrual. Payouts happen every two weeks, based on a statement the owner approves. Clients sign the offer and consents with a finger on a tablet and book free slots themselves in their account or via Telegram.

• A QR code for everyone who brings clients
• Each client stays tied to their source for good
• Discounts and earnings calculated automatically
• Checkout with shifts, a log and reconciliation
• Accounts for partners and ambassadors
• Payout statements approved by the owner
• Document signing on a tablet
• Booking and reminders in Telegram

Studio owner's review

The store sells car parts: ignition coils and spark plugs, wheel caps, emblems, tyre pressure sensors. While a buyer with a faulty part waited for an answer, they had time to request a return and post a negative review. Now there's a flyer with a QR code in the box: it leads to the store's page and from there to the AI assistant in a messenger.

The assistant understands which product it is and what happened, and answers from a knowledge base built from a year and a half of the store's conversations with buyers: it sends instructions, suggests a check or arranges a replacement. It collects the request details one item at a time, and an employee sees the whole conversation with photos in the dashboard and can take over at any moment.

• QR code in the box leads to a messenger
• AI recognises the product and the issue
• Answers only from the approved base
• Instructions, a check or a replacement
• Request details one item at a time
• Dashboard with conversations and photos
• Take over a conversation with one switch
• Two brands on one system

DS parts review