Telegram chats became a stable lead generation channel with a 5,000-ruble AI bot

Telegram chats can be more than just a place for casual conversation—they can be a working sales channel. The author of this case study shows how an agency built an AI bot that monitors open business chats, filters service requests, and delivers a steady stream of qualified leads for approximately 5000 rubles ($50–60 USD) per month.

Where to look for demand

The idea grew from a simple observation: in industry-specific Telegram chats, there are regular requests for contractors, advice, and help. For B2B teams, these are almost live bulletin boards where entrepreneurs, managers, and decision-makers communicate. The problem is that such requests get buried in thousands of replies.

Reading dozens of chats manually is expensive and time-consuming, and it doesn't make sense to hire a dedicated employee just for monitoring. That's why almost everyone has access to these channels, but few use them systematically. Instead of adopting a third-party service with unclear ROI, the team decided to test their hypothesis in-house.

Over one business day, a business analyst within the company built a bot that listens to open chats from a shared database, runs messages through a set of filters, and sends managers only those that look like genuine service requests. After that, a human decides whether to engage in dialogue and make an offer via direct message.

How selection works

The real value of the system turned out to be not just in reading chats, but in the logic for filtering out noise. Telegram has too much garbage: job postings, ads, small talk, jokes, and spam. So the bot works in several stages and doesn't try to dump everything into the CRM funnel. First, it cleans the stream, then checks the meaning of the text, and finally evaluates whether the message is actually a request for help rather than just a discussion of a topic.

"The bot only listens to open business chats from my personal database

and the database that employees built."

• An anti-spam filter removes messages with excessive emojis, advertising style, and template job listings.
• Lemmatization brings words to their base form and helps catch meaning even if the author writes with errors or jargon.
• A database of approximately 200 key lemmas connects messages to the agency's services: from performance marketing to audit and analytics.
• Intention analysis looks for phrases like "need," "looking for," and "recommend," to separate genuine demand from small talk.

After scoring, the message passes a final check through a prompt with a list of the agency's services. If the model considers the request relevant, the bot sends it to a corporate messenger, not directly to the CRM. This is an important detail: managers get a compact queue for manual validation and can change the logic without a developer if needed. According to the author, teams and prompts are stored in Google Sheets, so employees themselves add new services, key lemmas, and manage turning the bot on or off.

Economics and results

The project has an almost microservice-like budget. Initial development took about 100 dollars of business analyst work, the server costs approximately 5 dollars per month, and OpenAI expenses are 20–50 dollars per month. In total, that's roughly 5000 rubles in monthly costs. For comparison, a dedicated manager or external developer would have been significantly more expensive, especially considering the constant monitoring of tens of thousands of messages.

The bot processes tens of thousands of messages per month, leaving about 800–1000 after two levels of filtering, and then helps managers find 15–22 qualified requests monthly. Of the messages that made it to manual processing, 5–8% turn out to be relevant, and of those, 50–60% become leads for the agency. Because of the high volume of noise, the authors deliberately didn't connect the tool directly to the CRM—otherwise garbage would skew the statistics and waste sales team time.

What it means

The case shows that Telegram chats can become an inexpensive B2B lead generation channel if you don't try to automate everything to completion but leave the final decision to humans. This approach works best where customers hang out in thematic chats: marketing, advertising, real estate, consulting. The main asset here isn't the model itself, but properly configured filtering logic and a cheap operational setup that the team can maintain without constant developer involvement.