What "AI Agent" Actually Means

You've spent three levels working with tools that wait for you: you open the assistant, type a prompt, and read the answer. An AI agent flips that. It's given a goal and some tools of its own, and it decides what steps to take without you approving each one.

Concretely, an agent can read a webpage, run a search, send an email, or update a spreadsheet, then decide on its own what to do next based on what it found. A general assistant answers one question at a time. An agent chains several actions together toward a goal you set once.

A simple example: instead of you copying a customer email into ChatGPT and pasting the reply back, an agent reads the inbox itself, drafts a reply, and either sends it or drops it in a folder for you to approve, on its own, every time a new email arrives.

This is genuinely useful and genuinely riskier than anything in Levels 2 and 3. The tools you met there only act when you tell them to, one prompt at a time. An agent keeps acting after you walk away, which means a mistake can repeat itself before you notice.

That tradeoff, less supervision for more automation, is what the rest of this course is about: how to connect tools safely, where to start, and just as importantly, where not to hand over the wheel yet.