An AI agent is a component of a business process responsible for automatically conducting a dialogue with a user using artificial intelligence. Its primary task is to manage communication at a specific stage of the sales funnel: from the initial contact to the achievement of a specific goal (e.g., registration, purchase, or confirmation of an action).

Simply put, it is a customizable “smart conversationalist” that operates according to a predefined script.

🔑 The Essence of the AI Agent Block

The block combines:

• the communication goal (what needs to be obtained from the user),
• behavioral logic (how exactly the agent communicates),
• conditions for completing the stage (when to move on),
• additional materials (to enhance the dialogue).

⚙️How it works

1. You assign a task to the agent —describing the goal and communication style.
2. When a task is specified, the agent conducts the dialogue itself.
3. It relies on:
   • your instructions,
   • the context of the conversation,
   • attached materials (knowledge base).
4. When a transition condition is met, the user moves on to the next stage.

🎯 Main Function

Guide the user to the desired result without operator intervention, while maintaining a natural and controlled dialogue.

“General Settings” tab, field descriptions:
1) Name – Enter the internal name of the “AI Agent” block. This name will be visible only to you in the block diagram within the MVP system when configuring a business process. By default, the block is named “AI Agent.”

2) Delay - Set the delay between the next message. While the delay is active, all messages from the user will be ignored.

“AI Agent” tab, field descriptions:

1) From Template — add a task that was previously saved as a template.

2) Task for AI Agent – This is a key field where you define the agent’s goal and behavior scenario for this stage of the conversation. Here, you must describe exactly what the AI should achieve and how it should conduct the dialogue.

Recommendations:

• formulate the task clearly and concisely;
• use simple and unambiguous instructions;
• specify the communication style (e.g., friendly, businesslike, etc.);
• describe the desired outcome of the conversation.

📌 Example:
“Guide the customer through the registration process, respond in a friendly manner, explain the steps if necessary, and offer a bonus for the first deposit.”

3) Prompt Auto-Optimization — the system automatically analyzes performance statistics and improves the prompt once every 24 hours. If the new version shows worse results, the system automatically rolls back to the previous version. Version statistics can be viewed in the AI Conversions section.

In the field below, specify the number of dialogues required before automatic prompt optimization is triggered.

4) Condition for move to the next block – Describe the conditions under which the user is considered to have achieved the goal of the current stage, and the business process can proceed to the next step.

This could be:

• confirmation of an action by the user;
• a specific phrase or intent;
• completion of a target action (e.g., registration).

📌 Example:
“Proceed to the next block if the user has indicated registration.”

If no condition is specified in the “Condition for moving to the next block” field, the transition to the next stage will occur automatically immediately after the current task is completed.

5) Verify Goal Against CRM Record — when this toggle is enabled, the agent not only analyzes the client's responses but also verifies goal completion using data from the client's CRM record.

📌 Example:
A client states that they have made a deposit of $100. If the transition condition requires a deposit of this amount and CRM verification is enabled, the agent will monitor the "First Sales Amount" field in the player's CRM record for the corresponding value. Until the required amount appears in the CRM, the agent will continue the conversation and will not move the client to the next stage of the scenario.

If the toggle is disabled, the agent relies solely on the client's statements and does not verify information in the CRM. For example, if the client says they have completed registration, the agent will treat the condition as fulfilled and immediately proceed to the next step.

6) Knowledge Base Files (attach files) – add materials that will help the AI agent engage in more accurate and confident conversations with users. These files serve as additional context and can be incorporated into responses or sent to the user as needed.

You can upload:

• images (e.g., screenshots or instructions);
• video and photo reviews;
• voice messages;
• other supporting materials.

It is recommended that you explicitly specify in the agent’s task the situations in which these files should be used. For example:

• send feedback if the user is unsure;
• display a screenshot example when asking for confirmation of an action;
• share instructions if the user does not understand the next step.

This makes the dialogue more intuitive and persuasive, increasing the likelihood of the user completing the desired action.

Button functions when adding files:

• “One Time” button - enabled by default. The agent will send the file only once and will not resend it during the current session. Disable this option if the file needs to be sent multiple times. The setting resets when the agent is restarted.
• “AI Analysis” button - AI will automatically identify the file content and use its context during communication with the client. If you do not want AI to analyze the file automatically, add a manual description in the corresponding field. For example: “Review photo - send if the client has doubts.”

7) Side Task — configure actions that the AI will perform when specific mentions or events occur in a conversation with a client: filling in a field, adding a tag, creating a task, and more. 

When to Trigger — describe the condition or situation in which the task should be executed.

Available Actions:

• Fill Field
• Add Tags
• Remove Tags
• Add Note
• Create Task
• Change Stage
• Switch to Manual Mode

8) One-Time Task — enable this toggle if the associated task only needs to be completed once. After the task has been successfully completed, the agent will no longer check this condition or return to the task again.

*If multiple actions need to be triggered simultaneously when the conditions are met, click the “Add Action” button.

Block Operating Modes

The behavior of the block is determined by two main fields:

• Task for the AI Agent — defines what the agent should do at the current stage.
• Condition for Transition to the Next Block — defines the condition under which the business process should proceed to the next stage.

Depending on how these fields are configured, three operating modes are available:

1. Persistent

If both Task for the AI Agent and Condition for Transition to the Next Block are filled in, the agent will continue interacting with the user until the specified transition condition is met.

📌 Example:

Task: "Assist the user with registration and answer any questions that arise."

Transition Condition: "The user confirms that the registration has been completed."

In this mode, the agent continues the conversation and does not move the user to the next stage until the goal is achieved.

2. Fire-and-forget

If Task for the AI Agent is filled in and Condition for Transition to the Next Block is left empty, the agent will complete the task, send a single message to the user, and immediately move the business process to the next stage.

📌 Example:

Task: "Inform the user about the bonus available after registration."

Transition Condition: Not specified.

After sending the message, the agent will automatically complete the current stage.

3. Silent

If both Task for the AI Agent and Condition for Transition to the Next Block are left empty, the agent will not send any messages to the user.

However, all configured block actions will still be executed, including field updates, tag assignment, task creation, stage changes, and other automated actions.

📌 This mode is particularly useful for technical stages of a business process where actions need to be performed without any interaction with the user.

“Push Notifications” tab — field descriptions:

1) Name - specify an internal name for the prepared push notification. For example: Push1. This name is used for easier navigation and configuration.

2) Push Delay - set the delay before sending the next push notification. Each subsequent push will be sent after the last message in the conversation, according to the delay you specify.

3) AI Push - enable the toggle if you want the push text to be automatically generated using artificial intelligence. The system will independently create a message based on the dialogue context, taking into account the configured legend, instructions, and conditions, creating the impression of natural and meaningful communication.

4) Text - add a predefined message text.

5) Buttons - you can configure buttons attached to the message text or add them to media files (such as images or videos) even without accompanying text. Buttons allow the user to interact with the bot or scenario and trigger the required action when clicked.

6) Attached Files - add media files (photos, videos, video notes, voice messages, documents). Multiple files can be uploaded at once.

To add multiple push notifications to this block, click the “Add Push” button.

Once you have finished configuring the settings, click the “Save” button.