The ultimate goal of enterprise AI is not merely to answer questions when asked, but to autonomously monitor, detect, and respond to events as they occur i.e., proactive digital workers operating continuously alongside human teams. However, existing multi-agent systems are designed for request-response interactions, not continuous event-driven operation. Current work exhibits static task handling where plans execute to completion without handling new events, no preemption to interrupt workflows when priorities change, and naive arrival- order processing rather than business urgency prioritization. This paper presents an autonomous event-driven multi-agent orchestration architecture that can subscribe to multiple event channels and react dynamically to events through priority scheduling, preemption, and context switching. The system comprises a Task Manager for event-to-task conversion with priority backlog management, a Planner for two-phase agent discovery and plan generation, a quantum-bounded Executor with checkpoint-based replanning opportunities, and a Replanner for dynamic adaptation through preemption checking and plan modification. This enables workflow interruption and resumption, intelligent event merging into active tasks, and systematic replanning at checkpoints. Which are capabilities missing from current multi-agent architectures.
Dhanyamraju et al. (Thu,) studied this question.