Plan-and-execute pattern in Agentic AI - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding the Plan-and-Execute Pattern

In the plan-and-execute pattern, an AI agent first creates a plan and then executes it step-by-step. Which of the following best describes the main advantage of this pattern?

A It allows the agent to adapt its plan dynamically after each step without a fixed sequence.

B It guarantees the agent will always find the optimal solution without errors.

C It requires less computation because the agent skips planning and acts randomly.

D It separates decision-making from action, enabling clearer control and debugging of each phase.

💡 Hint

Think about how separating planning and execution helps in understanding and controlling the agent's behavior.

The plan-and-execute pattern divides the agent's work into two clear parts: first, making a plan, and second, carrying out that plan. This separation helps developers debug and improve each part independently.

❓ Predict Output

intermediate

Output of a Simple Plan-and-Execute Agent

Consider this simplified Python code of a plan-and-execute agent. What will be printed when running it?

Agentic AI

class Agent:
    def plan(self):
        return ['step1', 'step2', 'step3']

    def execute(self, plan):
        for step in plan:
            print(f'Executing {step}')

agent = Agent()
plan = agent.plan()
agent.execute(plan)

A Executing step3\nExecuting step2\nExecuting step1

B Executing step1\nExecuting step2\nExecuting step3

C No output because execute method is not called

D step1\nstep2\nstep3

💡 Hint

Look at the execute method and how it prints each step.

The execute method prints 'Executing ' followed by each step in the plan list in order.

❓ Model Choice

advanced

Choosing a Model for Plan Generation

You want to build an agent that uses the plan-and-execute pattern. Which model type is best suited for generating a detailed multi-step plan before execution?

A A sequence-to-sequence model that generates a list of steps as output.

B A simple linear regression model predicting a single output value.

C A clustering model that groups similar data points without outputting sequences.

D A k-nearest neighbors model that classifies inputs based on neighbors.

💡 Hint

Think about which model can produce ordered sequences as output.

Sequence-to-sequence models are designed to generate ordered sequences, making them ideal for creating multi-step plans.

❓ Hyperparameter

advanced

Hyperparameter Impact on Plan Quality

In a plan-and-execute agent using a neural network for planning, increasing which hyperparameter is most likely to improve the detail and length of generated plans?

A Increasing the maximum output sequence length during training and inference.

B Reducing the learning rate drastically to near zero.

C Decreasing the batch size to 1 for faster updates.

D Using fewer hidden layers to simplify the model.

💡 Hint

Consider what controls how long the generated plans can be.

The maximum output sequence length controls how many steps the model can generate in a plan. Increasing it allows longer, more detailed plans.

🔧 Debug

expert

Debugging a Plan-and-Execute Agent Failure

An agent using the plan-and-execute pattern sometimes fails because the execution phase tries to run steps not in the plan. Which code mistake below causes this issue?

class Agent:
    def plan(self):
        return ['step1', 'step2']

    def execute(self, plan):
        for step in ['step1', 'step2', 'step3']:
            print(f'Executing {step}')

agent = Agent()
plan = agent.plan()
agent.execute(plan)

A The execute method correctly uses the plan argument, so no error occurs.

B The plan method returns an incomplete list missing 'step3'.

C The execute method ignores the plan argument and uses a fixed list with an extra step.

D The agent never calls the execute method, so no steps run.

💡 Hint

Look at what the execute method loops over compared to the plan.

The execute method loops over a fixed list that includes 'step3', which is not in the plan. This causes execution of steps not planned.