Agentic AI

Agentic AI refers to AI systems that pursue goals by taking actions, not just generating text: they plan multi-step tasks, call tools and APIs, read and write real systems, and adjust based on results. The shift from answering to acting is what gives agents their value and their risk.

A chatbot's worst output is a wrong answer a human reads. An agent's worst output is a wrong action a system executes: a deleted table, a sent email, a committed change. That difference moves the security question from "what did the model say" to "what is the agent about to do," which is the domain of action governance.

Practically, an agentic system combines a model, a loop, and tools. The model reasons, the loop feeds results back, and the tools touch the world. Most security-relevant behavior happens at the tool boundary, which is why controls concentrate there.

Related

• What is action governance?
• The agent security stack
• All glossary terms