A Hierarchical Agentic AI Framework for Autonomous Production Task Orchestration and Resource Coordination in SAP Manufacturing Systems
Manufacturing execution in SAP-integrated enterprises is constrained by coupled machine capacities, material availability, maintenance windows, quality holds, and rapidly changing customer priorities. This paper proposes a hierarchical agentic artificial intelligence framework that coordinates production tasks through strategic, tactical, and operational decision layers while preserving enterprise resource planning (ERP) transaction integrity. A supervisory agent interprets enterprise objectives; planning agents allocate jobs and scarce resources; and execution agents negotiate machine assignments, monitor deviations, and request constrained replanning. The framework combines event-driven SAP integration, a shared manufacturing state graph, mixed-integer feasibility screening, multi-objective optimization, agent communication protocols, and human approval gates. To demonstrate how the architecture can be evaluated, a fully disclosed illustrative discrete-event simulation compares manual dispatch, rule-based scheduling, centralized optimization, and hierarchical agents across demand and disruption scenarios. Illustrative results indicate that the proposed approach can reduce average tardiness and improve utilization relative to selected baselines, although the values are synthetic and must not be interpreted as observations from an operational SAP deployment. We discuss reliability, explainability, security, implementation constraints, and validation pathways. The contribution is an auditable coordination design and a testable evaluation protocol rather than a claim of empirically verified industrial performance.
Authors
- Nagi Singh
- Suman Rana
Publication Details
- Journal
- Journal of Contemporary Science and Technology Management
- Published
- 2026-11-17
- DOI
- https://doi.org/10.66592/szsdr84a
- Primary Topic
- Scheduling and Optimization Algorithms
- Type
- article
- Field-Weighted Citation Impact
- 0.00