Custom CRM Telegram

A CRM built right into the Telegram deal chats for a real estate agency

The agency worked in Telegram chats while the CRM lived on the side and nobody kept it up to date. We built a CRM that reads those same chats: deals, stages, viewings, reports for managers via an AI assistant, and voice message transcription.

Client Real estate agency
Industry Real estate · rentals and sales
Duration 2 weeks to launch, then ongoing development
Type CRM inside Telegram
2 weeks from kickoff to switching the client bots
8 deal stages changed with one button in the chat
0 external services for voice transcription

The problem

The agency works with clients in Telegram: every deal has two chats, one with the client and one internal. Deals come in from the rental and purchase bots and from social networks. The head of the agency wanted the full picture: how many deals there are, which ones have stalled and with whom, who did what today. The cloud Bitrix24 "lived its own life": agents did not work in it, and its data did not match reality.

Before

  • The real work happened in chats, and deals reached Bitrix24 only as an afterthought.
  • Stages and owners were not updated: the "what is in progress" report was assembled by hand from correspondence.
  • Viewings for clients were agreed verbally, with no reminders.
  • Voice messages (and agents and clients mostly dictate) were never transcribed or searchable.
  • The source of a deal was lost, so it was unclear which advertising worked.

What we built

We did not make the team relearn anything. We built a CRM on top of the chats they already use. The service reads the deal chats (only those with a deal number in the title) and stores everything in a single agency database.

  • Deals from the bots. The rental and purchase bots create a deal, set up a working group and pin a panel with buttons in it. Messages from the client and the agent flow both ways.
  • Stages and client profile. Eight deal stages (selection, viewings, deal preparation, successful deal, failure with a reason, postponed demand and others) are switched with a button. The client profile covers budget, district, bedrooms and timing.
  • Viewings and reminders. A viewing is created in the bot and lands in a shared calendar, the agent gets a reminder, and the whole team receives a summary for tomorrow every morning.
  • Morning check-in and reports. In the working groups the bot asks "what is the status of this deal today?", managers get a report in the evening, a list of "missed" deals on Mondays, plus monthly conversion.
  • AI assistant for managers. Connected to the database: in plain language you can ask about deals, agent workload and who started work at what time.
  • Voice message transcription on our own server, with no external services; the text is visible in the bot and searchable.
  • A deal from a screenshot. The SMM specialist sends a screenshot of a social media conversation, the AI fills in the deal, and the manager reviews it.
  • Deal sources and commission tracking per transaction, contracts from templates, and soft deletion of deals.

Results

  • The team kept working in Telegram, but every deal is now in the database with a number, a stage, an owner and a history.
  • Managers receive reports on their own instead of asking agents for them.
  • Viewings are not lost: the calendar and reminders work automatically.
  • Voice messages turn into text and can be searched.
  • The data belongs to the agency: the database sits on a separate server with daily backups. The agency works without an external CRM and its subscription fee.
  • Development continues: selecting options from the property database and a chat with the client are being discussed as the next step.

The client's name is withheld under NDA. Real names and details are shared at the brief.

A similar task?

Describe it briefly — within 1–2 days we'll come back with an analysis and an hourly estimate. If a packaged solution fits, we'll say so honestly.

✈ Telegram