In plain terms
Generative AI produces something when asked: a draft, an image, an answer. Agentic AI is handed an outcome and works out the steps itself. The difference is like asking a colleague a question versus giving them a task and hearing back when it is finished.
Why it matters
It moves the value of AI from content to completed work, and it moves the management question from “is the answer good?” to “how much may this system do on its own?” Budgets, controls and accountability all have to be set for a system that acts.
Example
Month-end close: instead of answering questions about invoices, the system matches invoices to purchase orders, chases missing documents by email, posts the clean entries and leaves only the exceptions for the accountant.
Most often confused with
Agentic AI vs. Generative AI
Generative AI is the capability: producing text, images or code. Agentic AI is a way of using that capability, in a loop with tools and a goal. Every agentic system is built on a generative model, but most uses of generative AI are not agentic.
Under the hood
“Agentic” describes a degree, not a category. Useful dimensions: how many steps the system takes without a human, whether the model or code chooses the next step, which tools and permissions it holds, and how long it runs. Typical building blocks are a reasoning-capable model, tool use, memory, planning and an orchestration layer. The engineering work is mostly in evals, permissions and failure handling, since errors compound over long chains.