How do AI Agents execute tasks?
This is where things start getting even more interesting.
That first AI Agent I set up, in addition to writing articles (which we might think are just words), was also researching on the internet.
But how does the agent research on the internet?
That’s where tools come in.
We start to give agents access to the internet.
The first ChatGPT didn’t research.
It was trained on specific topics, and stayed limited to what it knew.
I even remember that when Pelé died, ChatGPT didn’t know.
And it didn’t know for days, until an update came.
It couldn’t research on the internet about that event.
Now, with AI Agents, artificial brains (LLMs) can use external tools.
It’s like you connect the brain to a tool.
It could be a tool to work on Twitter, to browse websites, to operate credit cards, to analyze data, to write code… and much more.
These are tools programmed to expand the agent’s capabilities.
Today, you can even plug an AI Agent into a robot, and have that robot walking around your house, with the LLM brain making continuous decisions.