3 min read

Setting up a Telegram AI agent for sales follow-up without writing code

Setting up a Telegram AI agent for sales follow-up without writing code

Last week a client asked me whether I could automate their Telegram sales follow-up

The request was simple: a prospect writes a question in the chat, the AI agent should answer within seconds, and if the prospect sounds interested, the agent should schedule a call. No developers. No API keys to manage. Just a working agent by end of day.

I used AgentForge to build that agent in under an hour. The web UI handled everything. Here is how it worked.

Background

The client runs a small B2B service. They have a Telegram channel where potential clients ask about pricing and availability. A human operator answered those questions manually, but response time was often three to four hours.

Prospects left the chat. Some never returned. The client was losing leads.

They did not have a developer on staff, and they did not want to learn Python or deal with cloud infrastructure. They just wanted a bot that sounded like them and answered correctly.

Problem

The core issue was speed. Manual follow-up could not keep up with the volume of incoming questions. The client tried a generic chatbot before, but it answered in a stiff, formal tone that did not match their brand. Prospects noticed and disengaged.

Cost was also a factor. Hiring a dedicated chat operator for a small business was not realistic. The client needed something that worked out of the box, with no code, and no monthly retainer for a developer.

Why it matters

Every hour of delay on a sales inquiry reduces the chance of closing by a measurable percentage. For this client, a three-hour delay meant losing roughly one out of every five leads. Over a quarter, that is a significant revenue gap.

Without a working agent, the client had two bad options: ignore the problem or overspend on staffing. Neither was sustainable.

Solution

I installed AgentForge with Docker Compose on a cheap VPS. The default stack includes the Hermes agent, a web UI, and a Telegram connector. No code changes were needed.

In the web UI, I created a new agent profile. The configuration form asked for three things: role, goal, and style.

  • 📋 Role: set to "sales assistant for a B2B service company".
  • 🎯 Goal: set to "answer pricing and availability questions, then qualify the lead and offer a call booking link".
  • 🎨 Style: set to "friendly, direct, short sentences, no jargon".

Then I wrote a basic behavior rule: if the prospect asks about pricing, the agent should list the three standard packages. If the prospect mentions a deadline, the agent should confirm urgency and offer a meeting. If the prospect does not respond within two messages, the agent should send one follow-up reminder and then stop.

The whole process took about fifteen minutes in the UI. No YAML files, no environment variables, no debugging of API calls.

After that, I connected the agent to the client's Telegram group. AgentForge generates a token automatically. I pasted it into the Telegram Bot API panel, and the agent went live.

The agent was handling real leads within one hour of starting the installation, and I did not write a single line of configuration code.

Result

Response time dropped from three hours to under five seconds. The agent answers consistently in the client's own voice. Prospects who engaged with the bot booked calls at roughly the same rate as human-handled conversations.

The client has not hired a chat operator. They review the agent's logs once per day and occasionally tweak the style or behavior rules through the web UI. No code, no overhead.

The agent runs on a $10/month VPS. That is the total cost of the operation.

They kept the follow-up reminder rule because it recaptured leads who went silent. About one in four of those reminders turned into a booked call. That was money they would have lost entirely before.

The client is now planning to add a second agent for internal operations — task reminders for their remote team. Same setup. Same web UI. No code required.

AgentForge

AI agents for real work. If you want to see how this works in practice, visit our product site and blog.

Share
Serhii

Serhii