Case studies
Work with numbers before and after
AI Sales Employee on 1C + Bitrix24
tasks from memory20 deals a dayAn autonomous agent reads 1C, finds debtors, dormant customers and stock shortages — and creates tasks for sales reps in Bitrix24 by itself, each with a ready-made call brief. The very first run surfaced 18 debtors and 143 short-stocked SKUs. 1C is read-only; writes to the CRM go through an allowlist of nine operations.
read the case study →Furniture manufacturing · Ufa · client projectMebelRoom — website, calculator and request tracking
1 page on a site builder132 pages + CRMA furniture manufacturer with revenue in the tens of millions a month was living with a one-page website. Now: a catalog of 100 real projects with filters, a calculator with a PDF quote, and every request automatically becomes a lead in Bitrix24 with its source and UTM tags. Plus a set of rules for the sales team: what to do with a request next. Launched in a month, domain migrated with zero downtime.
read the case study →Bitrix24 + Telegram · client projectAI Sales Supervisor
exports and pivot tablesa question in TelegramThe CEO talks to the CRM like a live analyst: pipeline, revenue by source, rep KPIs — the answer comes back as a finished report. The system watches unanswered leads and stalled deals on its own — 1,815 deals under control. The third system built for this client.
read the case study →Bitrix24 → Telegram · manufacturing“Production PM” — order tracking
a customer complaintan alert the day it stallsA system for a furniture manufacturer: it tracks order stages in Bitrix24 via REST and, when an order sits longer than the standard allows, notifies the person responsible in Telegram. Overdue orders surface the day they stall, not when an unhappy customer calls.
read the case study →Bitrix24 · data controlThe “Auditor” — data discipline in Bitrix24
manual CRM checksa daily report in TelegramA data-quality Telegram bot for a furniture company. Every day it checks deals and records — empty fields, broken statuses, stalled stages — and sends the people responsible a report. “Dead” deals have vanished from the CRM, and the manager sees how disciplined the database is without checking by hand.
read the case study →Multi-agent systemAn AI Agent Fleet runs the product on its own
2–3 days of manual worka 3-line report at 08:30Eight autonomous AI employees — an economist, a CRO, a data analyst, virality, customer support, a product strategist, a content marketer and a supervisor — track the product's metrics, run the unit economics and send the owner a short digest every morning. The whole fleet of 8 agents — 45,000 ₽/mo instead of an entire department's payroll.
read the case study →Own productA website as proof
telling the story in wordsa verifiable showcaseSelling development work rested on chat threads and retelling — before a call, a client couldn’t verify anything. Now everything can be checked before the conversation: case studies with before → after metrics, two pricing calculators, a live dialogue with an agent above the fold. Every edit reaches the production domain in 1–5 minutes with no human involved.
read the case study →Own productPersonal AI Knowledge Base
sorting notes by hand0 minutes a dayThoughts are written as a single stream into a notepad from a phone or a computer — no tags, folders or formatting. Twice a day an autonomous agent turns what was written into atomic notes with links and topic maps. The human confirms nothing — safety rests on technical safeguards, not on the owner’s vigilance.
read the case study →In-house product · e-commerceA factory for personalized children's books
three specialistsone questionnaireAn in-house product — this is how I test technology with my own money before putting it into your business. The child becomes the hero of the book, and their appearance stays consistent on every page. Text, illustrations and the print layout are assembled from a single Telegram questionnaire — no author, illustrator or layout designer. Files pass the print shop's validation automatically, and the print run arrives by courier.
read the case study →