AI Agentes

What are they and when should you use them?

AI Agents are ideal for complex tasks that require specific actions, such as gathering user data or interacting with external systems.

Recommended Use Cases:

• Structured information collection (name, email, CPF, etc.).
• Integrations with external APIs, queries, validations, and more.
• Processes like user registration, authentication, and customized inquiries.

How it works: The AI Agent operates autonomously to carry out specific tasks. It integrates with the main flow only after completing its work, ensuring modularity and maximum efficiency.

Not Recommended Use Cases:

• Complex negotiations or conflict resolution processes.
• Extensive, open-ended conversations without a clear, defined objective.
• Advanced support for sensitive situations or emotional containment scenarios.
• Critical or high-risk operations, such as loan simulations, handling sensitive financial data, or actions involving significant legal or financial consequences, as there can be a margin of error or deviation in AI responses.

What does delegating tasks to an AI Agent entail?

When using AI Agents in a conversational flow, it is important to understand that you delegate partial control over the system’s responses. Unlike traditional bots, which follow fixed rules and predefined answers, generative AI interprets available information and formulates original replies—even when handling queries that are vague or not explicitly covered in the knowledge base.

This means that:

• There will not always be complete certainty about the exact response the agent will generate.
• If the information in the knowledge base is insufficient, unclear, or outdated, AI might attempt to fill those gaps, producing answers that may differ from what is expected or desired.
• There is a natural variability in the way AI generates responses, even when correctly configured.

Therefore, it is essential to keep the knowledge base up-to-date, clear, and as complete as possible, while carefully considering this autonomy when designing critical processes or situations where accuracy and precise control are essential.

Bot identity

Defining your bot’s identity is essential to ensure the Artificial Intelligence understands the overall chatbot context. What does this involve?

• Be sure to include relevant information about the bot.
• Adjust the tone of voice according to the project's purpose and target audience.
• Achieve consistent communication aligned with your business objectives.

Practical Example:

Instruction:

You are a bot for a language academy specializing in providing information about Portuguese, English, French, and German courses. You maintain a friendly, clear, and helpful tone. Your goal is to guide users and efficiently answer their questions respectfully, always using simple and positive language. You guide the conversations step-by-step, ensuring a structured, informative, and pleasant experience for anyone inquiring about the courses.

Instructions

Instructions define the AI's behavior.

They must be clear, direct, and aligned with the task's objective.

Specify what data should be requested, how answers should be presented, and establish any necessary restrictions.

Practical Example:

Instruction:

#OBJECTIVE

Your goal is to assist the user in booking a language course offered by our company. You should guide them clearly, orderly, and friendly to gather all the required information and confirm their enrollment in the chosen course.

#STEPS TO FOLLOW

1. Ask the user for the first and last name and store it in ${fullName}.
2. Request their identification number (ID) and store it in ${courseID}.
3. Request their email address and store it in ${courseEmail}.
4. Ask the user which course they are interested in and store it in ${selectedCourse}.
5. Display a list with all the information collected from the user and ask them to confirm to complete the process. If the user chooses to modify their information, repeat the confirmation step again.
6. Wait until the user sends a message confirming their details are correct, then store their information with the action create record in Google Sheets. Thank the user for their information, inform them they will soon be contacted via email, and end the conversation politely.

#RULES

• Tone must be friendly.
• Include only one emoji per response.
• Responses must have a maximum of 25 words.
• Only accept responses related to data collection and language courses.

To learn more and discover best practices for creating instructions, access the following article: https://help.botmaker.com/en/help/6634855130683497129 

Natural Language with generative instructions

Expand your bot's ability to identify complex intents by interpreting the full context beyond single keywords.

Example: Set your Natural Language to recognize not just "products" or "company," but also the context of queries like "How many products does the company offer?" or "How long have you been in the market?"

Content Base

Allows faster loading and more precise bot responses by utilizing available knowledge bases. This update applies to Bot Designer with AI Agents and any Botmaker version.

Information quality and structure directly influence AI accuracy.

Para conocer más y mejores prácticas de la base de contenidos puedes leer el siguiente artículo https://help.botmaker.com/es/help/4174346223833880536

General Recommendations for Content Bases:

• Organize files by topic (e.g., User_Manual.pdf, Products.xlsx).
• Prioritize text-based files (text within images will not be read).
• For extensive tables or structured lists, use XLS/XLSX files with clear headings.
• For documents, prefer PDF/DOC/TXT formats with a maximum of 150 pages, and avoid long text blocks.
• Text within images: AI cannot read it. Always use plain text.
• Complex tables: Prefer XLS/XLSX files; avoid merged cells or complex formulas.
• Websites: AI reading capacity depends on site structure and internal organization (HTML, tags, accessibility, etc.).

Recommended Formats and Structure:

• Title (Heading 1): Mandatory, identifies the topic.
• Subtitle (Heading 2): Optional, adds detail.
• Detailed content: Use short paragraphs and bullet points for visual clarity.
• Divide content into main themes: Product manual, Rules, Processes, FAQs.
• Use descriptive, standardized names for files.

Practical Example: 

Portuguese Course

Overview: Course aimed at those looking to learn Portuguese from scratch or enhance their skills for travel, studies, or work.

Available Levels:

• Beginner (A1-A2)
• Intermediate (B1-B2)
• Advanced (C1)

Class Formats:

• In-person
• Live online (group)
• Private lessons

Duration and Frequency:

• Regular: 3 months (2 weekly classes, 90 minutes each)
• Intensive: 1 month (3 weekly classes)

Requirements:

• No prior knowledge needed (beginner level)
• Placement test available

Additional Information:

• Digital materials included
• Certificate awarded upon completion
• Limited spots (max. 8 participants per group)

Content Base Update:

Manual: Each time you update a file, you must upload it again to the platform.

Updated content will be available after closing the current conversation session.

Frequently Asked Questions (FAQs)

What is the difference between an AI Agent and Generative Response? An AI Agent executes tasks or completes flows by collecting data, while Generative Responses present existing information in a clear and understandable way.

For more information, we recommend the following article: https://help.botmaker.com/en/help/8783707342642706995

Should the knowledge base be updated manually? Yes, every time you update a document, you must upload the new version.

What should be done if the AI cannot find information in the knowledge base? Generative AI will attempt to produce answers, but these may not meet expectations if the content is unclear or unavailable.