Bots and AI agents for sales and support

Answer customers, take payments, and bring in a person when needed

Set up a conversation for your website and messaging channels. The agent answers from your knowledge base, checks customer details, and passes the conversation to a person when it needs to.

Build your first journey See how a journey runs

How a customer journey runs

Someone opens the chat on your website and asks whether a plan covers their team.
channel.website.message_received
The agent uses your knowledge base to answer and records the customer's request.
agent.intent = upgrade_question
Region, plan and history come from the unified customer profile. The branch is deterministic, so the same case always goes the same way.
profile.plan = team, region = EU
The journey creates a CRM lead and sends a payment link in the same conversation.
crm.lead_created, payment.link_sent
When confidence drops or the customer asks for a human, the conversation moves to your inbox with the profile, the variables and the reason attached.
handoff.reason = low_confidence
Conversion, resolution and revenue are linked to the step that produced them. Use that record to decide what to edit next.
outcome.paid = true

In production

How Gogol School answers students

Gogol School runs student sales and support on GetMyBot, across website chat and Telegram, with operators picking up the conversations the journey escalates.
Gogol School

Gogol School is a named customer whose case we publish with permission.

Read the case

What the agent can use

An AI agent with defined limits

The agent writes the reply. The journey controls its sources, actions, and handoff rules.

Answers from your content: Answers come from knowledge sources you upload and index. You can test retrieval before anything reaches a customer.
Actions you allow: The agent calls the tools you allow it to call, such as creating a CRM lead or checking a booking. Nothing outside that list is reachable.
Rules for handing over: Set the conditions for handing a conversation to a person, including low confidence or a direct request from the customer.
Check it before publishing: Run the agent against real conversations, score individual turns and review quality before you publish a new version.

How the agent is bounded See the journey builder

The condition panel of a journey step: two rules are enabled, plan equals pro and message count more than one, and a third rule sends the conversation to a named follow-up when they are not met — The rules that sit around the agent. Conditions combine with AND and OR, and the last one names exactly where a conversation goes when they do not match.

When the AI should stop

The operator gets the conversation history

The operator receives the conversation with the visitor's messages and the information collected by the journey.

Without a shared profile

The customer repeats the problem. The operator reconstructs the history from three tabs and answers from memory.

With the journey attached

The operator sees the full thread, the customer profile, the variables the journey collected and the reason it escalated. Routing rules, working hours and private notes decide who gets it.

An open conversation in the operator inbox, showing the customer name, an Open status, an Unassigned owner, a CRM link and the unread messages the visitor sent — The operator sees the conversation status, owner, and CRM record in the inbox.

See the shared inbox

Sales, support, operations

What would you like the bot to handle?

Answer the buying question, then take the payment

The agent qualifies interest, the journey checks the offer against the profile, and the payment link goes out inside the same conversation.

A customer asks
The AI agent reads intent
The journey checks the rules
The journey takes action
A person takes over
The outcome is recorded

What you measure Conversations that end in a paid order.

See how this is built

The analytics screen of a demo project for the last 30 days. A users card reads 240 with a badge showing 150 new, a payments card shows the amount recorded in the period, and a table lists the five reactions that ran most often with their trigger counts. — The real analytics screen, rendered over seeded demonstration data. It reports the population, the payments recorded in the period, and which automations actually ran. The numbers come from a demo project and are not a customer result.

Examples

Start with an example

Recommended start

Website question to paid order

Follow the website-to-payment journey shown above to set up a complete path.

Other starting points

Support desk with escalation
Use knowledge-base answers, routing rules, working hours, and a shared inbox for the questions that need an operator.
Bring quiet customers back
Behaviour-based segments and a follow-up sequence, measured against return rate instead of send volume.
Operations without the manual step
Use triggers, scheduled steps, and calls to your systems for recurring operational work.

Setup

Connect your channels and systems

Channels available today

Website chat
Telegram
WhatsApp
VK
Email
Instagram
Facebook Messenger

We list a channel here once it is generally available. Anything still in review is absent on purpose.

One customer history across channels: A visitor on your site and the same person in Telegram resolve to one profile, so history and attribution survive the channel switch.
Extend the flow with code: Use the REST API, scoped personal tokens, MCP, or your own WebSocket handler to extend the flow in code.
Manage your customer data: Export or delete customer data on request, keep an audit trail of every sync, and run on a dedicated server where that is the requirement.

Plans with included requests

Choose a monthly or yearly plan with a monthly request allowance. If you need more, buy extra requests separately.

All plans

Start with the question customers ask most often

Build the path from the first question to a measurable outcome, then add other paths as needed.

Build your first journey

Before you start

Which AI provider does it use?

You connect your own provider key, so the model choice and the bill stay with you. The journey logic around the agent is the same whichever provider you pick, though not every model supports every capability equally.

Can a person take over a conversation?

Yes. Operators work from a shared inbox with assignment, groups, routing rules, working hours, private notes and quick replies. The AI part stops where you tell it to stop.

Which channels can I use?

Website chat, Telegram, WhatsApp, VK, email, Instagram, and Facebook Messenger are available now. We list other channels when they become generally available.

Can I move an existing bot across?

Yes, there is an import path from SamBot that brings the scenario over rather than making you rebuild it. Complex logic still deserves a review after import.

How does pricing work?

A monthly plan with an included volume of processed requests. You pay for the volume you use, and the plan page shows exactly what each tier includes before you sign up.

Who owns the customer data?

You do. You can export or delete a customer record on request, and a dedicated or on-premise deployment is available when your policy requires the data to stay in your own infrastructure.