Some of the earlier posts in this series talked about AI assistants. These AI Assistants use their training (and sometimes extra data) to answer questions, summarise information, and do small tasks. They are good at producing human-like responses, but they are not truly independent. Usually they follow a simple loop: take input, process it, then produce output.

Agents are the next step. Instead of just answering, an agent can make decisions, use external tools, act on its environment, and check whether those actions worked i.e more autonomous. If something does not go well, it can re-iterate and improve the outcome.