Tool selection by the agent in Agentic AI - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

How does an agent decide which tool to use?

An agent has multiple tools available to solve a problem. What is the main factor it uses to select the best tool?

A The tool with the most complex internal architecture

B The tool with the fastest response time regardless of accuracy

C The tool that was used most recently

D The tool that best matches the current task requirements and context

💡 Hint

Think about what helps the agent solve the problem correctly, not just quickly or by habit.

An agent selects a tool based on how well the tool fits the current task and context to maximize success. Speed or complexity alone do not guarantee the best choice.

❓ Predict Output

intermediate

Output of tool selection code snippet

What is the output of this Python code simulating an agent choosing a tool?

Agentic AI

tools = {'search': 0.8, 'calculator': 0.9, 'translator': 0.7}
selected_tool = max(tools, key=tools.get)
print(selected_tool)

A "search"

B "translator"

C "calculator"

D KeyError

💡 Hint

Look for the tool with the highest value in the dictionary.

The max function with key=tools.get returns the key with the highest value. Calculator has 0.9, which is highest.

❓ Hyperparameter

advanced

Choosing the right confidence threshold for tool selection

An agent uses confidence scores to pick tools. What happens if the confidence threshold is set too high?

A The agent rarely selects any tool, possibly failing to act

B The agent selects tools more often, increasing speed but risking errors

C The agent ignores confidence and picks tools randomly

D The agent always picks the tool with the lowest confidence

💡 Hint

Think about what happens if the agent demands very high confidence before using a tool.

If the threshold is too high, the agent may find no tool confident enough and thus avoid selecting any, leading to inaction.

❓ Metrics

advanced

Evaluating agent tool selection accuracy

An agent selects tools for tasks. Which metric best measures how often the agent picks the correct tool?

A Accuracy

B Recall

C Precision

D F1 Score

💡 Hint

Consider the metric that measures overall correct selections out of all attempts.

Accuracy measures the proportion of correct tool selections out of all selections, making it suitable here.

🔧 Debug

expert

Debugging tool selection logic error

What error does this code produce when the agent tries to select a tool?

tools = {'search': 0.8, 'calculator': 0.9, 'translator': 0.7}
threshold = 0.95
selected_tool = None
for tool, score in tools.items():
    if score > threshold:
        selected_tool = tool
print(selected_tool.upper())

A KeyError because 'upper' is not a dict key

B AttributeError because selected_tool can be None

C TypeError because score is a float

D No error, prints 'CALCULATOR'

💡 Hint

What if no tool meets the threshold? What happens when calling upper() on None?

If no tool has score > 0.95, selected_tool stays None. Calling None.upper() raises AttributeError.