Autonomous Exception Management in SAP S/4HANA Manufacturing Through Multi-agent Generative AI and Event-driven Supply Networks

Thousands of system exceptions every day are created in manufacturing operations in SAP S/4HANA. Majority of these exceptions are being done manually. It is cumbersome, costly as well as inaccurate. This paper suggests a multi-agent generative AI combined with event-based signals of the supply network architecture as the way to make SAP S/4HANA a mechanized entity, automatically detecting exceptions, classifying their root causes, and running corrective actions. The suggested system relies on proven systems of AI applications in supply chain and operations management, design patterns of multi-agent systems (cyber-physical production) and available empirical findings concerning the benefits of AI-based operations and their limitations. Exception prioritization and load balancing of agents are based on two performance equations. Simulated findings of a discrete-event manufacturing situation demonstrate a decrease in the average time in which exceptions were resolved by 62% over a manual base. Paper provides the taxonomy of agent role, and a comparison of types of exceptions and strategies to resolve them.