What is Agentic AI?
AI systems that autonomously plan, invoke tools, observe results, and iterate toward a goal rather than answering a single prompt.
Definition
Agentic AI wraps an LLM in a loop: given a goal, the model plans, calls tools—shells, browsers, APIs, code editors—inspects the results, and decides the next step, continuing until the task completes or a guardrail halts it. The ReAct pattern (interleaved reasoning and action) underlies most implementations; frameworks include LangGraph, the OpenAI Agents SDK, and coding agents like Claude Code and GitHub Copilot's agent mode. Multi-agent setups delegate subtasks to specialized workers. Engineering effort concentrates on tool design, context management across long horizons, permission boundaries and sandboxing, and evaluation—since autonomy multiplies both usefulness and the blast radius of mistakes.