NL2ProcessOps: Towards LLM-Guided Code Generation for Process Execution

The optimization and automation of process deployment operations play a key role in enhancing efficiency and adaptability in Business Process Management. This paper presents NL2ProcessOps, a novel approach leveraging Large Language Models (LLMs) and concepts such as Retrieval Augmented Generation (RAG), agents, and tools for code generation to streamline process deployment operations. The proposed approach is designed to work with textual process descriptions and focuses on the various operations of process deployment, from extracting the control flow in terms of a process model, to retrieving required tools associated with each task, and generating executable code for manual refinement purposes and deployment in a process execution engine. The paper discusses the underlying principles of LLMs, the design and implementation of the approach, and its evaluation using a set of process descriptions. It demonstrates the effectiveness of NL2ProcessOps in generating high-quality code to support process deployment operations through both human and automated assessments. The paper concludes with a discussion of potential applications and future work.