The term “agentic AI” is used to describe a new class of AI systems that, unlike simple chatbots, can act independently and purposefully. These AI systems perceive data, think strategically, plan multi-stage processes and carry out actions - without constant human control. Agentic AI can therefore complete complex, multi-step tasks autonomously, for example in customer service or software development. The technological basis for this is a four-stage process:
1. perceive: information intake from databases, sensors or external tools.
2. reason (thinking): orchestrating problem solving through an LLM (Large Language Model) using retrieval techniques and other specialized models.
3. act: executing actions with tools or APIs.
4. learn: continuous self-optimization of the system through feedback loops.