What is an AI agent?

In 101 you built your first app. In 201 you turned it into a real tool: one that remembers data, recognises users and talks to a database. Impressive work. But in everything that tool does so far, you're still at the wheel. You click, you fill in, you start every step.

In this course we go beyond that step. We move from a tool that waits for you to software that gets to work by itself: an AI agent.

Before you build anything, it's important that you understand exactly what makes an agent different. That's what this lesson is about.

Three kinds of help, one big difference

The difference isn't in how clever the software seems, but in how much it does by itself. Compare these three:

• An ordinary app waits for you and does exactly one thing when you click. Your calculator from 101 adds up the moment you press the button, and not a second sooner.
• A chat with Claude gives answers and thinks along, but carries nothing out. You ask how to sort invoices, and you get a fine explanation back. You still do the sorting yourself afterwards.
• An agent is given a goal and works towards that goal by itself. It decides which steps are needed, uses tools, and keeps going until the task is done.

The recipe book and the cook

Picture a dinner. A recipe book tells you step by step what to do, but no meal appears on the table unless you go and chop and cook yourself. That's a chat: it knows everything, but carries nothing out.

A cook is only given the instruction 'make a dinner for four'. They decide themselves what to make, grab the pans themselves, taste along the way and adjust, and finally put a plate in front of you. That's an agent: you give a goal, not a list of clicks.

So an agent can take multiple steps in a row, make decisions along the way, and use tools (for example open a file or draft an email) without you steering each step separately.