# Multi-Agent System

> A multi-agent system is an AI application in which several specialized agents work together on a task, each with its own instructions, tools and responsibilities. A coordinating agent or a fixed workflow passes work between them, for example one agent researching, another drafting and a third reviewing. Frameworks such as LangGraph, CrewAI and Microsoft Agent Framework (the successor to AutoGen) are commonly used to build them.

## Why it matters for your business

Splitting a complex job across specialized agents can make each part easier to test and improve, and lets you give each agent only the data and permissions it needs. It also adds cost, latency and failure points, because every handoff is another model call that can go wrong. Many problems pitched as multi-agent work well with one agent and a good workflow. A multi-agent design earns its complexity when the subtasks are genuinely different, can run in parallel, or need separate permissions.

## Example

An insurance agency processes commercial renewals with three agents. The first extracts exposure data from the client's emailed documents, the second compares coverage options from carrier quotes, and the third drafts a renewal summary for the account manager. A workflow controls the order, and the account manager reviews the summary before anything goes to the client.

Related terms: https://www.lumeor.studio/glossary/ai-agent/, https://www.lumeor.studio/glossary/agentic-ai/, https://www.lumeor.studio/glossary/observability/, https://www.lumeor.studio/glossary/evals-ai-evaluation/, https://www.lumeor.studio/glossary/latency/

---

Source: https://www.lumeor.studio/glossary/multi-agent-system/