AI Agents

In Module 08, we learned that large language models (LLMs) are AI models that are trained to predict the next word and are therefore capable of generating text, right?

More recently, however, you have probably noticed that AI systems based on LLMs have acquired new capabilities and no longer just answer questions. They can plan tasks, ask clarifying questions, break goals down into steps, and produce concrete outputs such as code, reports, or files. Systems such as ChatGPT, Claude, and Gemini now go beyond passive interaction: they can execute tasks, make decisions, iteratively improve their performance, and increasingly interact autonomously with their environment. These systems are called AI agent systems.

AI agent systems can be seen as a more powerful version of generative AI systems. While generative AI systems create new content based on patterns learned during training, AI agent systems can also act actively, for example by searching the internet or databases, accessing calendars, classifying documents, preparing presentations, and much more. Most current AI agent systems are based on LLMs.