Omnichannel
WhatsApp, Instagram, Telegram. One agent.
One AI agent answers every channel from a single knowledge base, and one inbox holds the whole history. A lead can ask under an Instagram post, continue in WhatsApp and finish on a call without repeating themselves.
What runs on one agent.
WhatsApp Business
The official WhatsApp Business API, not a phone with a QR code. Incoming messages get an answer in seconds, and outbound campaigns run on approved templates instead of on a number that can be banned.
Instagram Direct
Direct messages, replies to stories and comments under posts. A public comment turns into a private conversation automatically, which is where most Instagram leads are otherwise lost.
Telegram
A bot that answers like a person rather than showing a menu of buttons. The same knowledge base, the same tone, the same write-back to your CRM.
Voice calls
The agent can call a lead who went quiet, qualify by phone and then continue the same conversation in chat. Chat and voice are one history, not two tools.
Website chat
An embeddable widget on your own site, answering from the same sources. A visitor who leaves a phone number can be picked up in WhatsApp later without losing what was already discussed.
One knowledge base for all of them
Prices, delivery, policies and product details are maintained once. Every channel answers from that one source, so customers cannot get two different prices from two different chats.
A shared inbox is not the same as a shared agent.
Most tools in this category unify the window a human looks at. The harder problem is unifying what the automation knows, and those are very different products.
| Multi-channel inbox with separate bots | One agent across channels | |
|---|---|---|
| Knowledge | Each channel has its own flow and its own answers | One knowledge base answers every channel |
| Context when a lead switches channel | Starts over, the customer repeats themselves | History carries across, the agent already knows |
| Changing a price | Edited in every flow, one gets missed | Edited once |
| Off-script questions | Falls out of the flow to a human or a dead end | Answered from the sources, escalated only when genuinely unknown |
| Voice | A separate vendor with a separate transcript | The same agent, the conversation continues in chat |
| What you maintain | One configuration per channel | One agent |
| Adding a channel | Build the flow again from scratch | Connect it, the knowledge is already there |
| Reporting | Message counts per tool | One funnel across every channel |
This is the whole argument for a single agent. It is not about seeing messages in one place, it is about not maintaining the same knowledge four times.
Each channel has its own rules.
Any honest omnichannel setup is shaped by platform policy, not by feature lists. These are the constraints that decide what automation can actually do for you.
WhatsApp has a reply window
After a customer writes, free-form replies are allowed for 24 hours. Outside that window you need a message template approved in advance, which is why speed of first answer matters commercially and not just cosmetically.
Instagram needs a professional account
Messaging automation runs on a business or creator account connected through Meta. The customer has to write first, so Direct is not a cold outreach channel.
Telegram is the most permissive
A bot can answer freely once a person starts the chat, which makes Telegram the easiest channel to connect and the least useful one for reaching people who have never contacted you.
Voice needs a reason to call
Outbound calling works on your own lists of people who asked to be contacted. The agent identifies itself as an assistant, and the recording stays with the conversation.
Templates are approved in advance
Anything you send first in WhatsApp has to be a template cleared beforehand. Campaign copy is therefore a planning job, not something written on the morning it goes out, and any tool promising unrestricted mass messaging is describing the unofficial route.
Sender reputation is shared across channels
Blocks and complaints lower the quality rating of a WhatsApp number and reduce how much you can send. Relevance is not politeness here, it is capacity, and it is why an agent that answers precisely beats one that answers a lot.
Why running three channels manually falls apart.
The failure is rarely the first message. It is the second one, in another channel, from a person nobody recognises as the same customer.
Traffic does not arrive in office hours
Instagram and WhatsApp peak in the evening. A morning reply competes against every other option the customer looked at overnight.
Nobody watches four windows
Comments, Direct, WhatsApp, Telegram and the site widget are five separate streams. Manual coverage means whichever one is open right now.
The same lead counted twice
Without one inbox the same person becomes two records and gets two different answers, which reads to the customer as a company that does not know its own prices.
Handover loses the thread
When a manager takes over from a bot in a separate tool, the customer explains everything again. That repetition is where people quietly decide to go elsewhere, and it never shows up in any report.
Nobody can see where deals die
With channels split across tools there is no single funnel to analyse. You end up with per-channel message counts and no answer to the only question that matters, which stage loses the money.
How it goes live.
Connect the channels you have
WhatsApp Business, Instagram, Telegram and the site widget connect one by one. You can start with a single channel and add the rest later without rebuilding anything.
Load the knowledge base once
Give it a website link or upload price lists and policies. The agent answers from those sources and says it will check rather than inventing a number.
Read the first transcripts
Correct tone and fill gaps in plain language, no prompt engineering. The correction applies to every channel at once because there is only one agent.
What being on one agent actually buys you.
The value is not the number of logos on a pricing page. It is what happens in the three or four moments where a normal setup drops the lead.
The channel switch survives
A person asks under a post, moves to Direct, then sends a number for WhatsApp. One agent keeps that as one conversation with one history.
Payment happens in the chat
The agent sends a payment link inside the conversation, with Kaspi as the first gateway live. The deal closes where the customer already is.
The CRM gets the result
amoCRM and Bitrix24 are native, so a lead becomes a deal with the transcript attached rather than a message somebody meant to write down.
Booking closes the loop
For services the agent writes the appointment straight into Altegio or Google Calendar, so the conversation ends in a slot rather than in an intention to call back and agree one.
One correction fixes every channel
Tell the agent in plain language what it got wrong and the fix applies in WhatsApp, Instagram, Telegram and on calls at once. With separate bots the same correction is four edits, and one of them is always forgotten.
What it plugs into.
An agent that answers well but writes nowhere still leaves your team doing the admin. These are the systems the result lands in.
amoCRM and Bitrix24, natively
The agent creates the lead, moves the deal and attaches the transcript, and it can read deal fields back so the answer takes the current stage into account. No middleware connector in between.
Booking and stock
Altegio and Google Calendar for appointments, MoySklad for availability. For services and retail this is what turns a good conversation into a slot or a reserved item rather than a promise.
Your own systems
Public API, webhooks, custom API tools and MCP servers, so the agent can check an order, a balance or a delivery status in whatever you already run. API access sits on the Pro plan.
How this compares to other omnichannel tools.
These products are not interchangeable, and the split that matters is aggregator versus agent. Several of the tools below are better than Pleep at the job they were built for.
- respond.io
- The strongest international omnichannel inbox in this list, with the widest channel coverage and the best documentation. Priced and supported for a global audience, with no Kazakh-language focus and no Kaspi.
- Umnico
- A mature messenger aggregator for the Russian-speaking market with a long integration list. Built inbox-first, so the AI is an addition rather than the product.
- Wazzup
- The default way to put WhatsApp and Instagram inside amoCRM or Bitrix24 in this region, and very good at exactly that. It is a connector, not an AI seller.
- Wazzapo
- Kazakhstan-focused WhatsApp Business API with visual workflows and AI replies. Strong on WhatsApp specifically, narrower across other channels.
- SendPulse
- A marketing suite where chatbots sit next to email and SMS. Attractive if you want campaigns and messaging from one vendor.
- ManyChat
- The best known Instagram and Facebook flow builder. Excellent when the journey is predictable and buttons are what you want.
- Conferbot
- An omnichannel chatbot builder covering many surfaces at once, aimed at teams that want to assemble flows themselves.
- Jirask
- A newer omnichannel assistant positioned around answering across messengers from one place. Worth a look if you want something lighter than a full platform.
- WiseChatt
- Another compact omnichannel option in this market, built around a single inbox with AI replies rather than around a builder.
- Pleep
- An AI sales agent rather than an aggregator: one brain across WhatsApp, Instagram, Telegram, voice and web chat, native amoCRM and Bitrix24, payment in chat through Kaspi, Russian and Kazakh as first-class languages.
| Question to ask | Why it decides the outcome |
|---|---|
| Is it one agent or one inbox? | An inbox unifies what a human sees, an agent unifies what the automation knows |
| Does the history follow a lead across channels? | This is the single most common place a multi-channel setup loses the deal |
| Is the WhatsApp connection official? | Unofficial automation risks the number, and the number is the asset |
| Which CRM is native, not via middleware? | A connector in the middle is one more thing to pay for and to debug |
| Can it take payment inside the chat? | Moving the customer to an invoice by email is where conversion drops |
| How is the Kazakh, tested on your own script? | Feature tables claim it far more often than live demos deliver it |
| Can the same agent make the call? | Splitting chat and voice across vendors splits the customer history too |
What it costs.
Pricing follows volume rather than seats, so connecting a fourth channel does not add a per-person bill.
Subscription by volume
A plan base plus a per-message rate. On the entry plan 1,500 messages a month comes to about $81.50. The calculator on the pricing page prices your own volume.
Voice billed per minute
Calls are charged on top by the minute, so the voice channel costs nothing while it is idle.
Or a share of revenue
There is a programme where you pay a percentage of the deals the agent closes instead of a subscription, for teams that would rather tie cost to outcome.
What you need to start.
Connecting channels is the quick part. What decides answer quality is whether your knowledge exists as text.
A WhatsApp Business number
A number that is not already tied to a personal WhatsApp account, plus a business that can be verified. This is the longest step and it is worth doing properly.
Prices and policies in writing
Anything the agent should answer has to exist as a document: a price list, a delivery policy, an FAQ, whatever you already send customers by hand.
Someone to read the first day
The first transcripts are where tone and gaps get fixed. After that it mostly runs itself, and corrections stay corrected.
Frequently Asked Questions
One agent for every channel you sell in.
Connect the channel that carries the most leads today and add the rest when you are ready.