Agentic AI
Chapter 8 · page 118
Definition
Agentic AI refers to artificial intelligence systems capable of acting semi-autonomously or autonomously to reach a goal, often over several steps. Unlike a standard chatbot, which merely answers an immediate request, an agentic AI can plan actions, use tools, access external services, remember certain information and adjust its behaviour according to the results it obtains. In practice, this autonomy rests on a language model orchestrated with instructions, a memory and tools. The whole then behaves like an agent carrying out tasks with little human supervision.
Example
An AI agent asked to organize a trip can search for flights, compare hotels, check the weather, book activities, fill in forms and send confirmations by email, all from an instruction such as: “Organize a weekend in Rome for two people in May.” Claude Code is an example of an advanced software agent. It can write code, modify files, run commands, call external tools and chain several actions together to reach a defined goal.
Why it matters
Agentic AI no longer merely generates content or answers questions: it can chain several actions together using tools and software. This capacity opens up significant possibilities for automation in work, services and research. But that autonomy also raises critical issues of safety, control and responsibility.
How to cite this text
Vial, S. (2026). Agentic AI. In A Living Lexicon of Artificial Intelligence (p. 118). Stéphane Vial, publisher, Montréal.
