What is Agentic AI?
Agentic AI describes systems where AI plans and carries out multi-step tasks on its own, using tools and feedback, instead of only answering one prompt at a time.
In more detail
The term names the design approach, while 'AI agent' names one piece of software. An agentic system can include one agent or several, plus the memory, tools and approval steps around them. What makes a system agentic is the loop: plan, act, observe, adjust. A chatbot that answers a question once is not agentic. A system that reads an email, opens the order in your ERP, notices the quantity is wrong and asks the buyer to confirm is. Vendors stretch the word, so ask what actions the system takes and who signs off on them. Autonomy is a dial, and we set it per step: low for anything that moves money or contacts a customer.
Example
A freight broker turns carrier emails into rate quotes with an agentic setup. One agent extracts lanes and prices, another checks them against contracted rates, and a dispatcher approves anything outside tolerance.