Connecting Knowledge and Tools

1. What Is Knowledge in an AI Agent?

Knowledge refers to the information an AI agent can use to answer questions and complete its assigned tasks.

This information may come from different sources, such as:

• Documents and PDF files.

• Websites and web pages.

• Company knowledge bases.

• Product catalogues.

• FAQs and user guides.

• Databases and structured information.

For example, an educational AI assistant may need access to course descriptions, lesson notes, learning materials and frequently asked questions.

Without relevant information, the agent may not be able to provide accurate answers about a particular course.

Example: AI Success Academy Learning Assistant

Suppose we want to build an assistant for AI Success Academy.

We could provide it with:

1. AI Foundations course information.

2. ChatGPT and AI Tools course details.

3. AI Agents course lessons.

4. Beginner AI learning resources.

5. Frequently asked questions.

The agent can then use this information to answer questions about the academy’s learning materials.

Important: Providing knowledge does not automatically guarantee accurate answers. The information should be relevant, up to date and properly organized.

2. What Are Tools in an AI Agent?

Tools are external capabilities that allow an AI agent to perform specific operations.

A basic chatbot mainly responds to user messages. An agent equipped with tools may also interact with other applications and services.

Examples include:

Tool	Possible Function
Web Search	Search for information online
Calculator	Perform mathematical calculations
Email Integration	Prepare or send emails with appropriate permission
Calendar	Check or manage scheduled events
Database	Retrieve or update stored information
File Reader	Access information from supported documents
API Integration	Communicate with external applications

For example, a travel assistant might use a search tool to find destinations, a weather service to check conditions and a booking integration to perform an authorized reservation.

The AI model decides when a connected tool is appropriate, according to its instructions and available permissions.

3. Knowledge vs Tools: Understanding the Difference

Knowledge and tools work together, but they have different purposes.

Knowledge	Tools
Provides information	Provides capabilities
Helps an agent understand	Helps an agent perform operations
May include documents and FAQs	May include APIs and integrations
Answers questions using available information	Searches, calculates or interacts with systems

Consider this example:

A customer asks:

“Is Product A available, and what is its price?”

The agent may use:

• Knowledge: Product descriptions and company policies.

• Tool: A live inventory database to check availability and current pricing.

The information and the tool work together to produce a useful answer.

4. How an AI Agent Uses Knowledge

There are several ways to make information available to an AI agent.

A. Instructions and Context

An agent can receive relevant information directly through its instructions or the context provided during a conversation.

Example:

“You are an educational assistant. AI Foundations is a beginner-friendly course covering nine lessons about artificial intelligence.”

This gives the agent some basic information.

However, large amounts of information should not necessarily be placed inside one instruction. Instructions should primarily define behaviour and responsibilities.

B. Documents and Knowledge Bases

Some AI platforms allow users to upload documents or connect knowledge sources.

For example:

• PDF learning guides.

• Word documents.

• Product information sheets.

• Course materials.

• Frequently asked questions.

The platform may retrieve relevant information from these sources when answering a question.

C. Databases

A database stores organized information that an agent can retrieve when needed.

For example, an online learning platform might maintain information about courses, lessons and student enrolments.

An agent can use an authorized database connection to retrieve relevant records.

D. Website Information

An agent may also use website content through an appropriate search, retrieval or integration system.

For example, a learning assistant could retrieve course information from the AI Success Academy website.

The method depends on the platform and its available integrations.

5. Understanding APIs and Integrations

You will often hear the terms API and integration when building AI agents.

An API, or Application Programming Interface, allows different software applications to communicate with each other.

Imagine two applications:

• Application A: An AI agent.

• Application B: A customer database.

An API provides a defined way for the agent to request information from the database.

For example:

1. A customer asks about an order.

2. The agent identifies the information required.

3. It uses an authorized API to request the order details.

4. The database returns the available information.

5. The agent explains the result to the customer.

This is one way AI agents interact with external systems.

You do not need to become a programmer to understand the basic concept. Many no-code and low-code platforms provide ready-made integrations.

6. Connecting Tools to an AI Agent

The exact process depends on the platform being used, but the general steps are similar.

Step 1: Identify the requirement

Decide what the agent needs to accomplish.

Example: A customer support agent must check order status.

Step 2: Select an appropriate tool

Choose a database connection, order management integration or API that provides access to order information.

Step 3: Configure the connection

Follow the platform’s instructions to connect the required application.

Step 4: Define permissions

Specify what the agent is allowed to do.

For example:

• Read order information.

• Search customer records.

• Prepare a support response.

Do not automatically grant permission to change or delete records.

Step 5: Test the connection

Ask the agent to retrieve sample information and check whether the result is correct.

Step 6: Monitor performance

Review how the agent uses the tool and correct problems when necessary.

7. Understanding Permissions and Safety

Connecting an agent to tools introduces additional responsibilities.

An agent that can access information or perform actions must operate within clearly defined permissions.

Consider these permission levels:

Permission	Example
Read	View a document or record
Search	Find information in a database
Create	Prepare a new support ticket
Update	Modify an existing record
Delete	Remove information

Not every agent needs all these permissions.

For example, an educational assistant may only need permission to read course information.

A customer support agent may need permission to create a support ticket, but not to delete customer records.

Important Safety Guidelines

1. Give agents only the permissions necessary for their tasks.

2. Protect passwords, API keys and confidential information.

3. Require human approval for sensitive or high-impact actions.

4. Test integrations before using them in real situations.

5. Keep connected information accurate and updated.

6. Monitor actions and review unexpected behaviour.

Remember: An agent should not have unlimited access simply because a tool is available.

8. Practical Example: AI Success Academy Learning Assistant

Let’s combine what we have learned.

Our proposed AI Success Academy Learning Assistant needs to answer questions about courses and guide beginners towards suitable learning materials.

Agent Role

An educational assistant for AI Success Academy.

Knowledge Sources

• AI Foundations course information.

• ChatGPT and AI Tools course details.

• AI Agents course information.

• Learning resources and FAQs.

Possible Tools

• Website search to locate relevant lessons.

• Document retrieval to find course information.

• A database connection to retrieve course records.

Permissions

• Read approved learning materials.

• Search course information.

• Recommend relevant lessons.

• Do not modify course records.

• Do not access private student information without authorization.

Example Conversation

Student: I am new to artificial intelligence. Which course should I start with?

AI Agent: You can begin with the AI Foundations course. It introduces basic AI concepts through beginner-friendly lessons. After completing it, you can explore ChatGPT and AI Tools to develop practical skills.

The agent uses its available course knowledge to guide the learner.

If connected to a suitable search tool, it could also locate the relevant course page and provide its link.

This is a simple example of combining a defined role, knowledge and tools.

9. Practical Exercise

Now design a simple AI agent that needs both knowledge and tools.

Your task: Create an AI Customer Support Assistant.

Answer the following questions:

1. What is the agent’s main role?

2. What information does it need?

3. Which documents or knowledge sources should be connected?

4. Which tools would help it perform its tasks?

5. What permissions should it have?

6. Which actions should require human approval?

Example Starting Point

Agent Name: Customer Support Assistant

Knowledge: Product information, FAQs and service policies.

Tools: Website search and customer support ticket integration.

Permissions: Read product information and create support tickets.

Boundary: Do not issue refunds or change customer account information without authorization.

Think carefully about the information and tools your agent actually needs.

10. Key Takeaways

In this lesson, we learned:

• Knowledge provides information to an AI agent.

• Tools allow agents to interact with external applications and services.

• APIs provide communication between different software systems.

• Documents, databases and websites can serve as knowledge sources.

• Tools must be selected according to the agent’s responsibilities.

• Permissions control what an agent can access or do.

• Security, testing and human oversight are essential.

Conclusion

An AI agent becomes more capable when it has access to relevant knowledge and appropriate tools.

However, connecting more tools does not automatically make an agent better.

The goal is to provide the right information, select useful tools and establish safe permissions.

Start small. Connect only what your agent needs, test its behaviour and improve it gradually.

In the next lesson, we will explore Designing Your First Agent Workflow and learn how to organize the steps an agent follows to complete a task.

Keep learning, keep experimenting and move one step closer to building your first practical AI agent!

AI Success Academy — Learn AI. Build Skills. Create Opportunities.

Lesson 5 | Course 4 | Lesson 7