Table of Contents

AI Agent Block

Jonathan Goodfellow Updated by Jonathan Goodfellow

What the AI Agent Block?

Welcome to the Brand New AI Agent Block! 🧱

AI Agents provide an intelligent conversational experience for your users. Based on your instructions, the AI Agent can collect and store information provided by the user, consult a Knowledge Base in order to answer questions, and seamlessly fit into your rule-based bot flows!

In this article, we will have a look at the Agent Blocks's main features and how they work βœ…

AI Agent Block Setup

Once you have added the AI Agent block to your flow, click on it to open the Setup Pop-Up:

Let's Dive into the features!

Build it for me

The Build it for me feature is an AI powered tool to help you create your Agent! You can either describe what you want the Agent to do in the Description section (you can use up to 1,000 characters), or you can choose from one of our Pre-Set Templates!

Once you have complete the description, just click Generate to have the AI complete the Agent Setup for you! πŸ˜‰

Using this feature will override any previous changes that you have made to the Agent Setup
Configuration

In the Configuration section you can set a Custom Error Message that will trigger in the event that there is a technical issue with the Agent (for example if there is an issue connecting with the AI tools in the back-end).

Knowledge Base

In this section, you can add all of the information that your agent will reference during chats with your users. Click the Add Knowledge button to upload a Knowledge Base for your Agent. You can choose to add Text, or upload a PDF File, or add a URL so that the Agent can scrape the website to gather information.

You can add 50,000 characters of text to your Knowledge Base directly, or upload a PDF which contains the text. For Web Scraping, the Agent can scrape up to 50,000 characters of the website's HTML code. The Knowledge Base can be trained on text only, and will not recognise information provided by images or graphics.

What to Include in the Knowledge Base:

  • Company or service-specific knowledge
  • Common FAQs and their answers
  • Any detailed information the agent should know about your business.
In the Knowledge Base you should not provide instructions for how your AI Agent should behave, just the information that it should consult.

The Agent uses this Knowledge Base when users ask questions or need information about your services. All uploaded content becomes part of the agent's understanding and response capability.

Agent Instructions

These instructions are the core of your AI Agent Setup. They define exactly how your agent should behave and operate πŸͺ„

Key Areas:

Here we will use the example of a Lead Generation Agent for an English Language Academy.

  • Behaviour Guidelines: How the agent should interact with users, for example:

Act as a Lead generation agent for Brighton English Academy

  • Response Style: How the agent should answer questions, for example:

Maintain a formal, polite professional tone throughout the chat

  • Edge Case Handling: Instructions for managing unusual situations, for example:

If the user asks about a course is not available, provide them with a list of the available alternatives taken from the Knowledge Base

  • Information Capture: Specific guidance on what data to gather and how, for example:

Gather the Following Information:

The course that the user is interested in (Business English, English for Beginners, Cambridge English)

This is directly related to the Information Capture section in the right side Panel. In the Information Capture section described below, you can set the Fields where the data should be saved and a description of the data. In the Instructions, you should instruct the AI Agent on which data should be collected and how.

You can find a full guide to creating instructions for you Agent with AI Tools, here.

Use Fields Button

With the Use Fields Button, you can make use of Fields that you have collected in your bot before triggering the AI Agent block.

For example, you might have collected the user's name earlier in the flow and now want the Agent to address the user by their name. You can select the Field from the drop-down list and add it to the Instructions!

Store Data

Here you can configure what information your agent should collect from users during conversations, and the Fields where the data will be stored.

Add Data

Click the Add Data button to create Fields. You should add the Field name where the data will be stored, as well as a Field Description giving a short explanation of exactly what is the data is βœ…

To ensure that this data is captured correctly, you should provide instructions to capture this data and give examples of the kind of data that the Agent should expect to collect here in the Instructions section.

For example, if you want the the Agent to ask the user's name and store in in a Field, it would look like this:

To ensure that this data is captured correctly, you should provide instructions to capture this data and give examples of the kind of data that the Agent should expect to collect here in the Instructions section.
Use Templates

With the Use Templates button, you can generate Fields and Descriptions from our pre-set options with templates for collecting Contact Information, Lead Qualification, and Real Estate information.

Your Agent can collect up to 20 Fields in this section

You can find a more detailed guide on how to set up Store Data feature in your AI Agent, here.

Outputs

AI Agents operate in a conversational loop. The user will remain in this loop until the Outputs that you define in this section tell the agent when to stop the conversation and trigger the next part of the flow.

For example, if you want the AI part of the chat to end when the user asks to speak to a Human, you would set it up like this:

When you return to your bot flow, you will see that the Output you created has been added to the AI Agent block, and you can connect this to the next part of the flow that you want to connect to when the Output is triggered.

In the "Speak to a Human" Output example, we could connect this directly to a Human Takeover block:

How It Works:

  • The agent continues conversing until specific conditions define in this section are met βœ…
  • Once conditions are satisfied, the conversation transfers to the designated block in your rule-based chatbot ➑️
  • This allows integration with APIs, CRM systems, or continued automated flows πŸ”—
The AI Agent does not have access to any context about what the bot being triggered when the conditions are met contains, so it will not understand instructions about what will happen next. For example, if in the next block you are triggering an email to be sent, you should not specify this in the Output, all you need to is instruct the chat to end.

And that's it! These 3 sections work together seamlessly to give your users an interactive conversational experience, tailored to yours and their needs! βœ…

Now that you learned about the main features of the Agent, we can move on to Testing and Troubleshooting!

Testing Your Setup

Once you have set up your Knowledge Base, Instructions and Outputs, we can move to Test your Agent to check that everything is working correctly!

You can test the flow in the same way you would any other bot, by Publishing it and Sharing with a Link from the Share section:

This way, you will not only be able to check that the Agent is answering questions correctly and accessing your Knowledge Base, but also that the Fields that you set to be capture by the Agent are being correctly saved. You can check this in the Analyze section of the Bot:

When testing, be sure to:

Verify Data Collection: Test that the Agent captures information exactly as specified in your instructions πŸ“ƒ

Monitor Agent Responses: Check responses to ensure that they comply with your Instructions and the information in your Knowledge Base πŸ‘¨πŸΌβ€πŸ’Ό

Check Outputs: Ensure the agent exits at the right time and transfers to the correct block in your rule-based flow ➑️

Test Integration: Verify that the collected information is properly accessible in your rule-based chatbot flow and any connected systems πŸ€–

Troubleshooting Common Issues

Even when following all the right steps and best practices, issues might still occur. Here is how to handle the most common ones πŸ”§

Agent Not Collecting Information

Check Your Instructions: Ensure your agent instructions include examples of how information should be collected and categorised βœ…

Verify Field Configuration: Make sure custom fields are properly defined and connected to your instructions βœ…

Outputs Not Working

Verify Block IDs: Double-check that the Outputs are connected to the right part of the flow in your Bot βœ…

Review Condition Logic: Ensure the Ouput Coditions are clearly defined and achievable based on your logic βœ…

Agent Not Using Knowledge Base

Content Format: Verify that the uploaded content is properly formatted and comprehensive enough for the agent to reference.

Instruction Clarity: Make sure your instructions guide the agent on when and how to use the knowledge base information.

And that's it! Be sure to check out our other AI Agent guides in our Knowledge Centre to keep learning! πŸͺ„

Was this article helpful?

How to Create an AI Agent to Detect User Intent

AI Agent Setup - Best Practices

Contact