4.1. Understanding AI Agents

💡 First Principle: A chatbot is like a reference librarian—it answers questions but can't do anything. An agent is like an executive assistant—it can check your calendar, book meetings, send emails, and decide what to do next based on results. The key difference is autonomy: agents have a loop (think → act → observe → repeat) while chatbots have a single turn (input → output).

What breaks without proper agent architecture understanding: Exam scenarios often describe complex workflows that candidates try to solve with elaborate prompt engineering or chained API calls. If you don't recognize the "agent pattern," you'll propose brittle solutions that require human orchestration of each step. Worse, you might implement a simple chatbot when the scenario clearly needs multi-step reasoning with tool access.

Consider this scenario: "A user asks 'What's the status of my order and when will it arrive?'" A chatbot can only respond based on its system prompt. An agent can actually look up the order by calling an order-tracking tool, interpret the response, and provide a real answer. The exam tests whether you recognize when a scenario requires this agentic capability versus when simple chat completions suffice.