Autonomous Event-Driven Multi-Agent Orchestration for Enterprise AI at Scale
The study compared two multi-agent orchestration methods, DAG Plan & Execute and ReAct, across various enterprise scales. It found that at large scales, performance is limited by agent discovery noise rather than task complexity. A new Task Manager component significantly improved handling continuous events and priority scheduling.