Researchers explore the use of LLM-based agents for RCA to address limitations in incident management, showing promising results in empirical evaluations.