Home Articles A robot in three channels

One sales robot for Telegram, WhatsApp and Avito

Technically yes: all three channels connect through AmoCRM, and the conversations land in one customer card. But it is not one button. In Telegram there is a difference between a personal number and a bot, the WhatsApp Business API needs preparation, and Avito is not built like a messenger and does not allow a robot everything.

In short: yes, but it is not one button

Technically one sales robot really can reply in Telegram, WhatsApp and on Avito at the same time. All three channels are connected through AmoCRM, and the conversations from them end up in one system. But “connect” does not mean “set up identically” — each channel has its own mechanics for receiving messages, its own limits and its own risks. The difference shows not in the robot's interface but in what happens between the client and the channel before the message even reaches the bot.

The easiest way to think about it is this: a channel is a pipe the correspondence flows through, and the robot is what answers at the other end. The pipe can be wide or narrow, official or a workaround, and it may allow the bot to write first or only to reply. Before promising a client “a single robot for all channels”, it is worth understanding what is hidden inside each pipe. Below, in order, channel by channel.

A personal number or a bot — the difference nobody mentions

In Telegram there are two different ways to be present: an ordinary personal account and an official bot registered through BotFather. With a personal number the robot looks like correspondence from a salesperson, but it has technical limits on automation and on who can write first. An official bot, on the contrary, is made precisely for automatic replies, but the client sees that it is not a person writing, and only the client can message a bot first — the bot cannot start a conversation itself.

In WhatsApp the story is similar, only stricter. There is ordinary WhatsApp, where the number is tied to a phone, and there is the WhatsApp Business API — the official interface for business through which automation works at platform level. The choice determines whether the robot can write to the client first, how many messages a day it may send and what happens if the client does not reply for a long time. These details are rarely discussed in advance, and then it turns out that a scenario which worked beautifully on paper runs into a limit of the channel itself.

For a client choosing how to connect, a simple conclusion follows: ask not “can WhatsApp be connected” but “through what exactly — an ordinary number or the Business API, and who writes first”. The answer to that question changes both how the pipeline is built and what can be automated in that channel at all.

Avito is built differently from messengers

Avito has no open public API for messaging that you can connect to directly, as you can to Telegram or the WhatsApp Business API. Access to the correspondence goes only through partner platforms that have their own agreement with Avito. That means the robot's functionality on Avito is limited by what the partner platform allows, not by what could theoretically be done with an open interface.

In practice this affects reply speed, the set of actions available inside a conversation and what client data reaches the CRM at all. The robot can reply, qualify and book a client — that is inherent to the task of a sales robot — but the scope of automation on Avito is usually smaller than in Telegram or WhatsApp, simply because the channel is technically more closed. That is not a reason to refuse to connect it, but it is worth knowing before the work starts rather than after.

Voice and photos deserve a separate mention: a sales robot can recognise voice messages and photographs, but on each platform this depends on what actually comes through the channel. If the Avito partner platform limits the types of attachment, recognition works only with what got through. So you should not expect identical capabilities in all channels — they are different by nature, and that is normal.

The WhatsApp Business API requires preparation, not just connecting

When people say “connect WhatsApp”, they often mean a one-step process. In fact the WhatsApp Business API requires business verification on Meta's side and the use of ready message templates for outbound messages if the client has not written first in the last twenty-four hours. That is a separate stage which happens before the robot starts replying at all, and it does not depend on how well the sales scenario is set up.

Templates have to be agreed with the platform in advance, and this concerns outbound messages specifically — a reminder about an abandoned enquiry, say, or a clarification about payment. Incoming messages from a client have no such limit: if the client wrote first, the robot replies freely within the open window. The difference between “replying” and “writing first” is as important here as in Telegram, only the rules are more formal and tied to Meta's moderation.

For a business this means that the time and scope of preparing the WhatsApp Business API depend on the company's specifics and on which messages its sales scenario needs. There is no universal “it takes so many days” here — too many variables sit on the platform's side. But knowing about verification and templates in advance is useful, so that it is no surprise when “just connect WhatsApp” turns into a project of its own.

One scenario or three different ones

A logical question: if there is one robot, does it work by one and the same scenario in all channels? Formally the sales scenario — the pipeline, the qualifying questions, the knowledge base about the product — can be a common core. But for each channel that core is usually adapted, because the format of communication differs: in Telegram and WhatsApp the correspondence is textual and often longer; on Avito the dialogue is shorter and tied to a specific listing.

Besides the format there are the channel's own limits discussed above: where the robot may write first and where it may only reply, what attachments are available, which actions the partner platform allows. All of this changes not the substance of the offer to the client but the way the robot phrases it and the order in which it asks questions. So the honest answer is: yes, the core of the scenario is common, but tuning the logic and the knowledge base for each channel is separate work, not copying one file into three places.

This is not complexity for its own sake but a consequence of the channels being physically different systems with different rules. A review of where money is actually lost in the current pipeline usually shows which channels to connect first and in what order — that is where any implementation of a sales robot begins.

What happens with an unofficial connection

The temptation to get around the limits is understandable: unofficial integrations with Avito or WhatsApp through an ordinary number sometimes promise more features and fewer formalities. But the platforms have mechanisms for detecting automated activity, and the result of an unofficial connection is the risk of the account being banned. For a business that means not just temporary inconvenience but the loss of a channel that enquiries were already coming through, and moving the number or account somewhere new from scratch.

The risk is especially tangible where the main flow of leads goes through the channel: if the account is blocked, the clients who wrote to that particular chat lose contact with the company until the number is restored or a new one is set up. Official routes — partner platforms for Avito, the Business API for WhatsApp — exist precisely so that automation does not look suspicious to the platform. That is not a guarantee of no problems at all, but a fundamentally different level of channel stability.

So when choosing how to connect a sales robot, it is sensible to ask not only “what can it do” but also “through which route of access to the channel does this work”. Saving at this step rarely pays off if the channel then has to be rebuilt from nothing.

How three channels converge into one client card

A separate task is not to duplicate a client who first wrote in Telegram, then on Avito about the same listing, and then called. amoCRM links conversations from different channels to the client card by the identifiers available — the phone number, the linked account, the data from the website form. This is set up when the CRM is assembled for a specific company's process; it does not work “out of the box” the same way for everyone.

The point is for the salesperson to see one history of correspondence rather than three disconnected conversations. The sales robot meanwhile remembers the conversation history within a single channel and hands a hot client to the salesperson with that context. But merging data from different channels into a single card is a matter of configuring the CRM and its pipelines, not a function of the bot.

That is exactly why connecting a robot rarely happens separately from setting up the CRM: without properly assembled pipelines and connected channels, duplicate leads are almost inevitable. A good way to check how a particular contractor does this is to ask not about the fact of connecting a channel but about the logic of merging cards inside amoCRM.

Where connecting really begins

The work begins not with setting up the bot but with reviewing the client's current sales department: which channels already exist, where enquiries are lost, what happens to messages in the evening and at weekends. On one live project, GetGate, it turned out that 27% of enquiries arrive exactly in the evening and at weekends — they used to be answered only the next morning, and that time was genuinely lost. Such a review determines which channels to connect first.

After the review the CRM is assembled, the robot is trained for the client's niche — for their specific questions, products and services — and only then the channels themselves are connected: Telegram, WhatsApp, Avito, the website. On the same GetGate project the assistant counts 5 product groups automatically, and that became possible precisely because the knowledge base and the logic were set up for that specific range rather than taken as a universal template.

Then the robot replies and books clients while the salespeople keep control of the hot enquiries — at GetGate this freed up about an hour of a salesperson's working time a day. Separately it was checked how comparable the assistant's answers are with a human's at all: in a blind comparison on real correspondence, in 80% of cases the assistant's reply was no worse. These are the figures of one live project, not a market average, but they show that the quality of the reply and the correct setup of the channels are matters of the same order.

What to bear in mind before connecting three channels

Coming back to the original question — can one sales robot be connected to Telegram, WhatsApp and Avito at once — the answer is yes, with a caveat: these are three different technical routes to one core of logic. Each channel has its own rules on who writes first, its own limits on automation and its own risks with an unofficial connection. Understanding these differences before the work starts saves time on rework afterwards.

A review of the specific situation usually shows which channels are worth connecting right away and which after the main flow is running. Sales analytics and quality control help to see where the robot copes well and where the scenario needs tuning for a particular channel. This is an ongoing process, not a one-off installation.

If it is not clear right now which channel to start with and what exactly happens to enquiries between Telegram, WhatsApp and Avito, the sensible first step is to review the current pipeline. After that decisions are made on facts rather than on general expectations of the word “automation”.

Frequently asked questions.

Are an AI agent and a chatbot the same thing?

No. A chatbot is given a script, an agent is given a goal. The bot follows a branch drawn in advance and gets lost at any step aside; the agent understands what is written in words and chooses what to do next itself.

Will the client realise they are not talking to a person?

Most likely yes, and passing the agent off as a human is not worth it: it comes out quickly and spoils the impression. In practice people are annoyed not by a robot but by silence.

What if the agent answers wrongly?

A properly configured agent answers from your materials and hands everything it doesn't know to a person. For the first weeks after launch you have to read real conversations and refine the answers — without that step the launch cannot be considered finished.

Does an AI agent need a CRM?

Formally no, it works without one. But then most of the value is lost: the correspondence stays in the messenger, no history accumulates and there is nothing to measure the result with.

Will an AI agent replace salespeople?

No, and there is usually no such goal. It takes over the first contact and the routine so that a person can deal with what needs a person: negotiation, difficult cases and money.

How much does an AI agent cost?

The price depends on the number of channels, the complexity of the scenario and the integrations needed. A sensible order of operations is to first count how many enquiries are being lost now, and decide on payback with that figure in hand.

See how it works.

Write a couple of lines about your business to our agent — it will reply and book you a free review. That is the demonstration.