Use what you learned in the previous lesson to solve real-world problems.

Check what you understood with a short quiz.

Decide what facts an agent needs before acting: the user’s intent, current situation, preferences, relevant history, and constraints. Practice leaving out background that does not change the next decision.

Use what you learned in the previous lesson to solve real-world problems.

List the actions an agent is allowed to take, such as search, ask a question, draft a message, book a slot, or hand off to a person. Reason through how the action menu shapes what the agent can and cannot accomplish.

Use what you learned in the previous lesson to solve real-world problems.

Check what you understood with a short quiz.

Read feedback as evidence about whether an action helped, failed, or changed the situation. Distinguish clear signals, like “payment declined,” from weak signals, like a vague user reply.

Use what you learned in the previous lesson to solve real-world problems.

Use feedback to choose the next move instead of blindly continuing the original plan. Practice when an agent should retry, switch actions, ask for clarification, or stop.

Use what you learned in the previous lesson to solve real-world problems.

Check what you understood with a short quiz.

Compare a chatbot that answers a prompt with an agent that keeps pursuing a goal across steps. Identify when conversation is just the interface and when the system is actually deciding and acting.

Use what you learned in the previous lesson to solve real-world problems.

Compare a fixed script, a rules-based automation, and an agent handling the same task. Notice how agents use feedback to choose among actions, while scripts and simple automations mostly follow a preset path.

Use what you learned in the previous lesson to solve real-world problems.

Check what you understood with a short quiz.

Set boundaries such as time limits, spending caps, tool permissions, data access, and stop conditions. Reason through how limits make an agent safer and easier to trust without making it useless.

Use what you learned in the previous lesson to solve real-world problems.

Identify moments where a human should decide: ambiguous values, high-stakes outcomes, irreversible actions, or cases where the agent lacks enough context. Decide when the agent should recommend, ask approval, or hand off entirely.

Use what you learned in the previous lesson to solve real-world problems.

Check what you understood with a short quiz.

Review this chapter with practice based on your mistakes.