Explainable Agentic AI for Predictive Autosys & Cybermation Job Orchestration

This research introduces an explainable agentic AI framework for predictive job orchestration in enterprise workload automation systems, focusing on Autosys and Cybermation environments. Traditional job schedulers lack adaptive intelligence and transparency, resulting in downtime, manual overrides, and operational inefficiencies. Our proposed system leverages reinforcement learning agents and causal inference models to proactively detect job failures, optimize schedule dependencies, and autonomously adjust execution paths. The framework includes an explainability layer powered by SHAP and counterfactual reasoning, enabling compliance with IT audit requirements and operator trust. Experiments on synthetic enterprise workloads demonstrate a 91.3% reduction in job failure rates and 87% improvement in SLA adherence. The solution offers a path toward self-healing, audit-compliant workload orchestration in regulated industries