The agent does instead of only telling. It is a language model handed hands and a goal, and that one change turns the model you chat with into something that goes off and works.
Last of the three quick looks, and the one that breaks the pattern. The chatbot talks. The image generator makes. Both of them hand you something and stop. An agent does not stop. It is worth seeing why, because this is where AI stops feeling like a tool you operate and starts feeling like something you delegate to.
A model with hands and a goal
Underneath, an agent is a language model, the same next-word predictor from a couple of pieces back. The difference is what it is allowed to do. Instead of answering your question and going quiet, it is given a goal and a set of tools, other software it can reach, and it takes steps toward that goal. It acts, looks at what happened, corrects, and goes again, over and over, until the job is done or it gets stuck. The chatbot tells you how to handle a job. The agent tries to get the job done itself.
That gap is the whole point of the agent. Ask a chatbot to sort out a messy folder and it writes you careful instructions. Hand the job to an agent and it goes into the folder and moves the files. I use one called Claude Code to write and edit the code behind this site, and it works the way I just described, I hand it a goal and it does the steps, opening files, making changes, checking its own work, coming back when it is done or when it needs me.
More power, more room to go wrong
The catch rides in on the same thing that makes it useful. A chatbot that is confidently wrong hands you a bad sentence, and you catch it. An agent that is confidently wrong takes a bad action, and by the time you look it may have taken ten more on top of it. It does not just say the mistake, it runs with it. That is not a reason to keep away from agents. It is the reason you stay in the loop with one, watching, keeping a hand on the wheel, giving it room to work without handing over the keys entirely.
The core of it
You do not need the deep version of this yet, and there is a deep version, agents are moving fast and they earn a project of their own later. For now, remember one line. An agent is an AI that acts instead of only answering, which makes it the most useful and the most worth watching of the three. That is how you spot one the moment it shows up in something you use.
That closes the map and the three tools on it. You have seen what AI is, where it came from, the kinds there are, and each of the big three up close. From here the series turns. We stop asking what these things are and start asking what they are worth to you, beginning with how far one can take you once you learn to guide it.
Next up: What It Can Do When You Guide It →