4 min read
Last week, someone asked Claude to plan a weekend trip. Not just suggest ideas. Actually plan it. Find flights. Check hotel availability. Look at restaurant reviews. Put it all in a calendar.
It did.
That's not a chatbot. That's an agent.
A chatbot answers questions. That's what most people have been doing with AI up until pretty recently. You type, it responds, you type again. It's a conversation. A very smart, very fast conversation.
An agent does things.
It can search the web, open apps, read documents, send emails, write code, fill out forms. Not all agents do all of those things, but the basic idea is the same: instead of just talking to you, it can reach out and interact with the world on your behalf.
The shorthand people use is "tools." An agent is an AI that has been given tools to use. The ability to search. The ability to write a file. The ability to click a button on a website. Each tool is something it can do, not just say.
Think about the difference between a notepad and a personal assistant.
A notepad is incredibly useful. You can write anything in it. But it just sits there. It holds information and gives it back to you when you need it.
A personal assistant can go do things. "Find me three plumbers with good reviews who are available Thursday. Actually book the one with the best reviews if they're under $150 an hour."
You don't have to be involved in every step. You say what you need and come back when it's done.
That's where AI has been heading. From very smart notepad to something that can actually go run the errand.
None of this was possible with the older versions of AI you might have tried. Early AI models were only good at generating text. They could tell you the steps for booking a flight. They couldn't book one.