Multi-agent

Multi-agent systems leverage autonomous AI entities to decompose complex workflows into specialized, collaborative tasks.

Multi-agent orchestration moves beyond single-prompt limitations by deploying specialized LLM instances (like those in Microsoft’s AutoGen or OpenAI’s Swarm) to solve high-stakes problems. By assigning distinct roles—such as a 'Coder' to write Python script and a 'Reviewer' to debug it—these systems reduce hallucination rates and handle long-running operations. Companies use this architecture to automate software engineering, market research, and supply chain logistics (where one agent monitors inventory while another negotiates vendor contracts). This modular approach ensures that if one agent fails, the entire pipeline doesn't collapse: it simply iterates.

https://www.microsoft.com/en-us/research/project/autogen/