Lesson 3

Tool Impact

More tools = different behavior

The Problem: AI without tools is like a person without hands. It knows things but can't do anything with that knowledge. It can describe how to check the stock price, but can't actually check it.

The Solution: Give AI the Right Tools

The right tools transform what AI can do. It's not about making AI smarter -- it's about giving it the ability to act. Through function calling the model gains access to external services, and with the ReAct pattern it can decide which tool to use at each step.

Think of it like a person with and without a smartphone:

  • 1. Without tools: "I think the capital of Mongolia is..." / "The weather might be..." / "Let me calculate in my head..."
  • 2. With tools: "The capital is Ulaanbaatar" (Google) / "It's 23C and sunny" (Weather app) / "The answer is 847.32" (Calculator)

Types of Tool Impact

  • Knowledge tools: Search, databases, APIs — extend what AI knows
  • Action tools: Send emails, create files, control devices — let AI do things
  • Computation tools: Code execution, math — precise calculations
  • Memory tools: Save/retrieve info — persistent context

Fun Fact: Studies show that giving AI access to a simple calculator reduces math errors by over 90%! The AI doesn't get "smarter" — it just stops pretending to be a calculator and uses a real one instead.

Try It Yourself!

Use the interactive comparison below to see how the same AI performs with different tool sets. Notice how dramatically its capabilities change!

⚙️How Tools Affect AI Behavior

🎯 AI agent behavior directly depends on available tools. Same question — different answers!

Enabled tools:(0/4)
Try different questions:
Without tools
Question:
What's the dollar exchange rate today?
🤷 "Sorry, I can't give you the current exchange rate as my data might be outdated. I recommend checking a bank website."
With tools
Question:
What's the dollar exchange rate today?
👆 Enable tools above
Enable matching tools
Key Insight

The set of available tools defines what an agent can do. Without search, it can't get current data. Without a calculator, it might make mistakes. When designing an agent, think: what tools does it need for the task?

Try it yourself

Interactive demo of this technique

Technique Comparison
Demo Mode
Pre-recorded responses
TaskBeginnerAnalysis

When was the latest SpaceX spacecraft launched and what was its mission?

Without technique
Without technique
Prompt
When was the latest SpaceX spacecraft launched and what was its mission?
Response

The latest SpaceX launch I know about is the Crew-6 mission in March 2023. It delivered astronauts to the ISS.

Tokens:30/35
Time:380ms
Quality:
With Tool Impact
With technique
Prompt
Answer using the search tool for current data. Tools: search(query) Question: When was the latest SpaceX spacecraft launched and what was its mission?
Response

Action: search("latest SpaceX launch 2025") Observation: SpaceX launched Starlink Group 12-5 on February 2, 2025 from Cape Canaveral. A Falcon 9 rocket deployed 23 Starlink satellites to orbit.

Answer: The latest SpaceX launch was the Starlink Group 12-5 mission on February 2, 2025. A Falcon 9 rocket deployed 23 Starlink satellites to low Earth orbit from Cape Canaveral.

🧠Question about a current event — model training data may be outdated
👁️Action: search() — retrieving current launch data
👁️With search tool: exact date, mission name, and details
Tools dramatically improve accuracy for current-event questions
Tokens:58/112
Time:2200ms
Quality:
Why this works

Without tools the model gives outdated training data. With search — up-to-date information. This is the key argument for tools in factual queries.

1 / 2

Lesson Quiz

1 of 3

1.How do tools extend the capabilities of LLM agents?

Practice Challenges

Create a free account to solve challenges

3 AI-verified challenges for this lesson

Related lessons:Function CallingReact

This lesson is part of a structured LLM course.

My Learning Path