Why do DAG-based workflow engines struggle with AI agent work?

DAG-based workflow engines are poorly suited for AI agent work due to their core design constraints. They are built for fixed, sequential deterministic tasks, while agents often operate with long lifespans, non-deterministic operational steps, and need critical state retention across unexpected interruptions these tools do not natively support. These tools remain ideal for straightforward scheduled work, and Catalyst fixes these key gaps by decoupling steps from rigid DAG structures.