Aug 2026· Process Science· Vol 3· 0 citations· 29 references
TL;DR
The results show that the applied LLM can reliably detect structural and semantic differences between formal business process models using Business Process Model and Notation, while distinguishing them from acceptable variations, demonstrating strong potential for automated model validation.
Abstract
Large Language Models (LLMs) are emerging as a promising tool in Business Process Management for comparing and validating process models. In this study, we evaluated an LLM’s ability to compare reference models with systematically modified variants representing typical modeling mistakes as well as harmless variations, such as layout or wording changes. The results show that the applied LLM can reliably detect structural and semantic differences between formal business process models using Business Process Model and Notation, while distinguishing them from acceptable variations, demonstrating strong potential for automated model validation. However, the LLM’s performance and accuracy are influenced by factors such as model complexity, the number of inserted modifications, and the total number of models and modifications provided simultaneously. High reliability is achieved when models are presented in a standardized, semi-structured format and supported by clear prompting instructions. Even multiple models can be processed effectively, up to a certain threshold of total modifications. Overall, the findings suggest that generative Artificial Intelligence tools for natural language processing, such as LLMs, may provide meaningful support in process model validation, offering efficiency gains and a level of abstraction that exceeds manual comparison.
The automated generation of business process models from natural language descriptions has recently attracted growing attention at the intersection of Business Process Management (BPM), Natural Language Processing (NLP), and Large Language Models (LLMs). This paper presents a Systematic Literature Review (SLR) on the current state of research in this emerging field. Following the guidelines of Kitchenham et al. and the PRISMA framework, 29 studies published between January 1st, 2023 and March 10th, 2026 were identified, selected, and analyzed. The review addresses two research questions focusing on the applied methodological approaches, the used LLMs, the employed modeling languages, as well as the evaluation strategies, challenges, and limitations reported in the literature. The results show a clear shift from traditional NLP-based techniques toward LLM-only and hybrid approaches. OpenAI’s GPT family, especially GPT-4 and its variants, dominates the field, while BPMN is by far the most frequently used target process modeling language. Furthermore, existing studies evaluate automated process model generation primarily through output-focused methods, such as quantitative metrics, expert reviews, and comparisons with alternative or human-created process models. At the same time, the reviewed studies reveal important challenges, including the continued need for human involvement and the output quality. Overall, current approaches show strong potential, but they still act more as intelligent assistants than as fully autonomous process modelers.
L. F. Hörner, Maximilian Möller, Manfred Reichert· IEEE Access· 0 citations
The proposed model aims to support the formalization of model selection processes, improve decision-making, and enhance the traceability and transparency of LLMOps practices and forms part of a broader research effort toward the formalization of the entire LLMOps life cycle.
Maria Chernigovskaya, A. Nahhas, Christian Haertel et al.· Proceedings of the 21st Inte...· 0 citations
The paper addresses the use of large language models (LLMs) to automatically generate executable business processes in BPMN 2.0 from unstructured natural-language descriptions, with deployment to process engines such as Camunda Platform in mind. The text-to-BPMN XML mapping task is stated and decomposed into subproblems: extracting activities, events, and gateways; recovering control flow and branching; ensuring valid sequence flows and conformance to the BPMN specification. We survey process representations (BPMN XML, JSON as an intermediate format, graph-based models) and LLM adaptation methods: prompt engineering, instruction tuning, and fine-tuning. We argue that edge (sequence flow) generation is the weakest link once nodes are fixed, and summarize typical structural failure modes (dangling nodes, disconnects, gateway violations, etc.) and causes tied to autoregressive generation. A staged pipeline is proposed—separate generation of node set V and edge set E followed by post-validation — together with a three-level validation scheme: syntactic (BPMN XSD), structural (graph invariants), and executable (Camunda deploy and run). The article outlines a feedback-enabled pipeline architecture and discusses applicability and limitations.
Gennady G. Bulgakov, S. Yarushev· SOFT MEASUREMENTS AND COMPUT...· 0 citations
This survey provides the first systematic synthesis of LLM-based diagram modelling research, highlighting needs for standardised benchmarks, stronger evaluation protocols, broader diagram coverage, and techniques for improving semantic reliability and multi-view consistency.
The results show that appropriate model slicing significantly improves completion correctness while simultaneously reducing token usage across several structural and semantic evaluation metrics, and establish model slicing as a key factor in LLM-based model completion and provide guidance for effective context selection for other modeling tasks.
A hybrid framework that takes a BPMN process model and a security requirements document as input and automatically generates security annotations adhering to the SecBPMN2 specification is presented, providing a scalable foundation for security-by-design BPM.