Large Language Model support in drafting a Business Process

Business Process Management has a powerful worldwide interest for years, and according to some studies, the investments into the field are continuing to grow.

Lykovas, Andrejus, 2025

Art der Arbeit Master Thesis
Auftraggebende
Betreuende Dozierende Jüngling, Stephan
Views: 31 - Downloads: 7
According to the 2018 survey, BPM has been very high on most companies’ list of priorities since the survey started back in 2003. Paul Harmon’s (2018) survey indicates that organisations go through cycles of interest in the Business Process Management work. Each of the peaks of interest are preceded by some novel concept, new technological development or economical cycle, even if the overall interest keeps growing (Harmon, 2018) With the advent of relatively cheap and readily available LLMs, we are on the potential verge of the new BPM hype cycle. Researchers and companies alike are exploring all possible applications of the LLMs in the daily operations to gain efficiencies, improve profits and simply gain a competitive advantage over the competition. Business process modelling is the key capability in any BPM initiative. It helps to ensure the quality of the communication between the various stakeholders, structures the common business understanding and streamlines the ease of understanding between different domains such as business and IT. No BPM initiative could be successful without the modelling capabilities, as is a crucial part in every step of the Dumas et al., 2018 BPM lifecycle.
This thesis picked up where several researchers such as Friedrich et al., 2011, Bellan et al., 2020, and Grohs et al., 2023 and others have left off and tackled the crucially important BPM lifecycle step of process discovery using novel Large Language Models. The design science research strategy was used when implementing and evaluating several prompt engineering techniques to develop a prototype. The designed prompt prototype was used to interview four process participants to evaluate the capabilities of Anthropic’s Claude 3.5 Sonnet and OpenAI’s ChatGPT-4o to generate an XML BPMN code to represent a process.
The paper contributes further knowledge and delivers a LLM prompt prototype. The prototype supports the process analyst by guiding the process participants through the interview discovery process, and generating a BPMN business process draft.
Studiengang: Business Information Systems (Master)
Keywords
Vertraulichkeit: öffentlich
Art der Arbeit
Master Thesis
Autorinnen und Autoren
Lykovas, Andrejus
Betreuende Dozierende
Jüngling, Stephan
Publikationsjahr
2025
Sprache der Arbeit
Englisch
Vertraulichkeit
öffentlich
Studiengang
Business Information Systems (Master)
Standort Studiengang
Olten