AI agents are software components that use AI (often an LLM) to decide what to do next and then take action. Unlike a simple chatbot that only returns text, an agent can call external tools: search the web, run code, query a database, or call an API. That makes them suitable for automation: booking a meeting, summarizing a document and sending it somewhere, or guiding a user through a multi-step process.

Agents are given a set of tools (functions or APIs they can call). The model reasons about the user's goal, chooses which tool to use and with what inputs, and interprets the result. That loop—think, act, observe—can repeat until the task is done. Workflows orchestrate these steps: sometimes a single agent runs the loop; sometimes a coordinator delegates to specialist agents. Frameworks like LangChain, LangGraph, and others help you define tools, prompt the model, and handle the execution loop.

Use agents when the task requires multiple steps, external data, or decisions that depend on context. Simple one-shot questions are often better handled by a single API call. For deeper coverage of how agentic systems work, their advantages, and governance, see the Agentic AI concept in this path.