Agentic AI describes systems that pursue a goal on their own in several steps, use tools and check intermediate results, instead of just answering one question.
Example
Instead of just listing flight times, the AI compares offers, checks your calendar and prepares a booking for your approval.
For pros
Typical building blocks are a language model, tools (tool calling, MCP), memory and an agent loop. The more autonomy, the more important human approval, narrow permissions and logs become. “Agentic” is also often used as a marketing word.
