Optimizing organizational efficiency with AI
Improving business process management with AI and automation
November 14, 2025
As operations scale, process models multiply. They get copied, tweaked, and scattered across teams-slowing work and increasing risk. PhD researcher Mahdi Saeedi Nikoo built AI-driven methods that help organizations cut duplication, improve consistency, and complete models faster.
His thesis, defended on Thursday, November 13, 2025, shows how to detect "clones" across large repositories and how to recommend the missing steps in incomplete subprocesses. The goal: save time, reduce errors, and speed up delivery across complex workflows.
Why this matters for management
- Reuse what works: Surface and standardize process models that already perform well.
- Cut waste: Eliminate duplicate models and rework across teams and business units.
- Faster delivery: Use smart suggestions to complete subprocesses without waiting on experts.
- Lower risk: Improve consistency and compliance by aligning on shared patterns.
What the research covered
Service composition languages. Saeedi Nikoo analyzed 14 languages for coordinating services across systems. The most relevant approaches today rely on central orchestration or a mix of central and distributed coordination. Many earlier options have faded out, which helps teams focus on what's viable now.
Clone detection in process models. Large repositories often contain duplicate models and repeated fragments. The proposed method identifies both full-model and fragment-level clones so teams can retire copies, reuse proven parts, and keep processes aligned. Tests showed it can outperform or complement leading tools such as Apromore, depending on the case.
Insights from open-source repositories. A large-scale study across sources like GitHub revealed frequent duplication at both model and fragment level. The data shows how models are reused, adapted, and spread across domains-useful for setting governance and reuse strategies.
Intelligent recommendations for incomplete subprocesses. The work compares similarity-based tools with large language models (LLMs). Similarity methods excel on smaller pieces; LLMs do better on larger, complex fragments. A hybrid approach gives modelers the most reliable suggestions and speeds up design.
How to apply this in your organization
- Inventory your models: Map where process models live, who owns them, and how they're used.
- Deploy clone detection: Identify duplicate models and fragments; retire copies and promote a single source of truth.
- Create a reference library: Curate reusable subprocesses and patterns; version them and set approval rules.
- Pilot hybrid recommendations: Use similarity for quick wins on small gaps and LLMs for complex sections.
- Set reuse KPIs: Track time-to-model, error rates, and the percentage of designs built from approved fragments.
- Close the loop: Feed accepted suggestions back into the library to improve future recommendations.
Standards, tools, and further reading
- BPMN Specification (OMG) - widely used for modeling business processes.
- Apromore - a well-known platform for process mining and analysis.
Upskill your team
If you're rolling out AI and automation across operations, equip managers and analysts with focused training. Explore role-based options here: AI courses by job.
About the researcher
PhD researcher: Mahdi Saeedi Nikoo, Department of Mathematics and Computer Science
Thesis title: Supporting business process management: clone detection and recommendation techniques
Supervisors: Mark van den Brand, Önder Babur, Sangeeth Kochanthara
Your membership also unlocks:
