Modeling meets Large Language Models

Modeling business processes is often challenging due to its complexity and potential for errors. One key issue arises when process experts and modelers are different individuals, which can lead to communication gaps and result in low-quality business process models. Recognizing this, our paper prioritizes the initial phase of modeling in Business Process Management (BPM). We propose a method that leverages Large Language Models (LLMs) to efficiently transform written business process descriptions into comprehensive graphical models. This approach offers a standardized and streamlined procedure to enhance the quality and effectiveness of business process modeling. While we focus on Petri nets as a primary example, our approach is adaptable to other graphical modeling languages. We present a novel method involving a series of LLMs to extract essential data, setting the stage for creating various graphical models. This technique aims to generate initial drafts that can be further refined, and its sequential application allows for adaptability to different modeling tools, including but not limited to the Horus Business Modeler.