Multi agent

Multi agent is a method where multiple AI agents share roles and collaborate. It divides the reviewer, writer, and verifier.

Multi-agent is a method of forming multiple AI agents with different roles like a team to solve problems, instead of leaving all tasks to one AI. For example, when writing a report, the agent who researches the data, the agent who writes the draft, and the agent who reviews errors collaborate in a relay.

It emerged because of the limitation that mistakes can easily accumulate if one AI handles a long and complex task alone. Splitting roles allows each agent to focus on his or her own role, which is used for complex tasks such as coding automation or in-depth research.

However, increasing the number of agents does not automatically improve the results. As agents exchange incorrect information, errors are amplified, and side effects such as costs and time are multiplied.

✅ Why it matters

• Accuracy can be increased by dividing complex tasks into roles
• Review and verification steps can be added to filter out single AI mistakes
• The design method is intuitive, modeled after the human team collaboration structure

⚠️ Limits and debates

• As the number of agents increases, costs and response times increase significantly
• Errors can be amplified due to incorrect information being passed between agents
• In some cases, it is worse than a single agent due to excessive design for simple tasks.