A PhD researcher from the Eindhoven University of Technology has demonstrated how artificial intelligence (AI) and automation enhance business process management (BPM).
As organizations rely more on business process models, managing large collections becomes challenging. Models can be duplicated, inconsistent, or fragmented, making them harder to reuse and slowing down operations
Mahdi Saeedi Nikoo tackled these challenges by creating techniques to identify duplicated process models across large repositories and deliver intelligent recommendations for incomplete subprocesses. His work integrates empirical studies, automated tooling, and AI-driven methods to make process modeling more practical and efficient.
Evaluating modern service composition languages
Modern organizations increasingly depend on orchestrated services to deliver complex functionality. Service composition languages enable them to define how various services – such as software components or business processes – interact to accomplish sophisticated tasks.
Saeedi Nikoo examined 14 such languages, assessing their features, adoption levels, and shifts in popularity over time. His findings reveal that today’s most prevalent approaches either manage service coordination from a central point or blend centralized control with more flexible, distributed setups. Many earlier languages have fallen out of use.
These insights help organizations and researchers identify which modeling approaches remain most relevant in the current landscape.
Register for All Access: Future of BPM 2026!
Identifying process model duplications
Managing large collections of business process models is challenging because similar or identical models often appear repeatedly. To tackle this issue, Saeedi Nikoo developed a method for detecting duplication (also known as clones) within repositories. His approach identifies both fully repeated models and smaller fragments that recur across different processes.
By uncovering these duplicates, organizations can cut down on redundancy, reuse existing models more effectively, and maintain greater consistency in their process landscape. Experiments showed that his method can outperform or complement leading tools like Apromore, depending on the context.
Insights from open-source repositories
Saeedi Nikoo also conducted the first large-scale analysis of business process models in open-source repositories, including GitHub. He investigated their domains, the tools used to create them, patterns of ownership, and the prevalence of duplication. The study found that clones are widespread at the model and fragment levels, highlighting how models are reused and adapted across different contexts.
These results offer researchers and practitioners practical guidance for managing large process model collections more effectively.
Join us at All Access: AI in PEX 2026!
Intelligent recommendations support modelers
Finally, building on the above findings, Saeedi Nikoo created intelligent recommendation systems to help modelers fill in missing steps in incomplete subprocesses. Comparing traditional similarity-based tools with advanced large language models (LLMs), he found that similarity tools work best for smaller pieces, while LLMs are more effective for larger, complex fragments.