A context graph is a unified, evolving graph that combines enterprise knowledge (people, docs, tickets, systems) with time-ordered action/event traces (create, comment, escalate, resolve) across tools, turning messy real-world execution into probabilistic workflow paths so AI agents can understand how work actually gets done, predict what step comes next, handle deviations, and learn continuously from both human and agent runs rather than relying on static records or hard-coded processes.