Studies on the Use of Large Language Models for the Automation of Business Processes in Enterprise Resource Planning Systems

Business processes in companies today are characterized by increasing digitalization and automation through the use of information systems such as enterprise resource planning (ERP) systems. To relieve the users of information systems of manual, repetitive and error-prone activities, a variety of process automation tools are available. However, applying such tools requires comprehensive process knowledge and an in-depth understanding of their concepts, implementation and use. Furthermore, previous automation tools lack in flexibility to react on changing conditions or process errors. This paper introduces an approach that applies large language models (LLM) to automate business processes in ERP systems to overcome the disadvantages of previous approaches. As SAP is the world's leading provider of ERP software, their ERP system was chosen to investigate the extent to which business process knowledge can be embodied in LLMs. It is examined how the possibility of natural language interaction provided by LLMs can be used to enhance the automation of the procure-to-pay process in SAP systems, as one core process and how process automation can be made stable against errors. In addition, the multi-agent framework AutoGen is used to simulate separation of duties (a concept for ensuring security in IT systems) and thus to investigate the involvement of multiple users in a business process.