The AI Assistant is an add-on service that is activated separately from the basic Message.Help plan. Before setting it up, make sure of two things: you already have a working project in the system, and the communication channels (WhatsApp, Telegram, and others) through which the assistant will communicate with customers are connected.
If the AI Assistant option isn’t displayed in your plan, contact support to have it activated.
Important: Each response from the AI assistant uses up requests from your request package. You can purchase a request package in the “My Plan” → “Plan Settings” → “Services” section. We recommend estimating the expected workload for the assistant in advance and monitoring your remaining requests so that the bot doesn’t stop responding to customers at the worst possible moment.
Next, you’ll need to go through four configuration sections, which together determine the bot’s behavior and logic.
This is where you lay the foundation for the assistant’s operation—who it is, who it communicates with, and when it’s active.
At this stage, you determine the specific purpose of the assistant—your choice will determine the set of additional fields.
If the goal is sales, you’ll need to specify the desired outcome of the conversation (for example, “sign up for a trial session”), the typical length of the sales cycle, a link to the online registration form, and decide whether the bot should sign up the customer itself or simply forward their contact information to a manager.
If the goal is to sign up customers, simply provide a link to the online sign-up form.
If the goal is technical support, you need to describe what constitutes a successfully resolved inquiry, list typical customer issues, and decide whether to ask the customer to rate the quality of support after the issue is resolved.
Regardless of the goal you choose, you can also configure the bot’s communication style here:
Based on the settings from the previous blocks, the system automatically generates a script—a step-by-step sequence of how the bot conducts a conversation with the client.
You can and should edit the script manually: refine the wording, add new steps, and tailor the logic to the specifics of your business. After making changes, it’s recommended to regenerate the working prompt—this will allow the bot to update its behavior in accordance with the new script.
⚠️ Try not to edit the prompt directly, bypassing the structured script—this can disrupt the context and lead to unpredictable responses from the bot. The correct approach: change the script → regenerate the prompt.
To ensure the assistant provides customers with accurate and helpful responses, it needs to be “trained”—that is, provided with information about the company. There are several ways to do this, including simultaneously:
⚠️ Only factual information should be entered into the knowledge base: services, prices, hours of operation, addresses, contact information, and booking rules. Do not include the bot’s internal behavior algorithms here (for example, “if a customer is upset, forward the conversation to a manager”)—the assistant will interpret such an instruction as a fact about the company rather than a behavioral rule, which will lead to errors in the responses.
Once you’ve filled out all the sections and saved your settings, the AI assistant is ready to go. You can return to the configuration at any time via “Settings” → “AI Bot Settings.”
The assistant’s behavior depends entirely on the settings you specify—technical support does not adjust the bot’s logic, as this is a tool configured by the business itself. If the assistant responds differently than expected, look for the cause in three places: the dialogue scenarios, the knowledge base, and whether the prompt is still relevant after changes.
Don’t expect perfect results on the first try—configuring an AI assistant is an iterative process. The more accurately and thoroughly you fill out the dialogue scenarios and knowledge base, the more confidently the bot will conduct the conversation.
Before connecting the assistant to real customers, be sure to test it yourself: ask various questions, simulate complex situations, and check how logically and correctly it responds.