[Paper Review] Recent Advances in Data-Driven Business Process Management

Lars Ackermann, Martin Käppel+22 more|arXiv (Cornell University)|Jun 3, 2024

Business Process Modeling and AnalysisBusiness, Management and Accounting3 citations

TL;DR

This position paper proposes a research agenda for data-driven business process management (BPM), emphasizing the integration of emerging technologies like generative AI, large language models (LLMs), and advanced data analytics to enhance process transparency, automation, and compliance. It introduces a 'predict and comply' pipeline for real-time compliance monitoring, leveraging predictive process mining and AI-driven techniques to improve decision-making, data quality, and process automation across the BPM lifecycle.

ABSTRACT

The rapid development of cutting-edge technologies, the increasing volume of data and also the availability and processability of new types of data sources has led to a paradigm shift in data-based management and decision-making. Since business processes are at the core of organizational work, these developments heavily impact BPM as a crucial success factor for organizations. In view of this emerging potential, data-driven business process management has become a relevant and vibrant research area. Given the complexity and interdisciplinarity of the research field, this position paper therefore presents research insights regarding data-driven BPM.

Motivation & Objective

• To address the paradigm shift toward data-driven decision-making in business process management (BPM) driven by the explosion of data and AI advancements.
• To identify and prioritize key research areas that bridge technological potential with practical BPM applications across the process lifecycle.
• To strengthen the foundation of BPM through improved data quality, novel data sources (e.g., unstructured data), and AI-enhanced analytics.
• To promote process-aware automation by integrating predictive monitoring, compliance prediction, and AI-based recommendations.
• To foster interdisciplinary collaboration between academia and industry to advance evidence-based, reliable, and transparent BPM systems.

Proposed method

• Proposes a 'predict and comply' pipeline that reverses traditional compliance monitoring by first predicting process outcomes and compliance states before runtime enforcement.
• Leverages predictive business process monitoring techniques using event logs and real-time streams to forecast KPIs and compliance risks.
• Integrates large language models (LLMs) and NLP techniques to extract insights from unstructured data such as emails, videos, and audio for process modeling.
• Applies model-checking techniques to verify process models against compliance rules, both ex post (from logs) and online (from streams).
• Explores advanced AI techniques such as transfer learning, federated learning, and neuro-symbolic AI to improve robustness and generalization in data-driven BPM.
• Develops frameworks for measuring and improving process data quality and provides AI-based recommendations for process implementation and optimization.

Experimental results

Research questions

• RQ1 How can predictive compliance monitoring be enhanced through a 'predict and comply' pipeline that integrates predictive process monitoring with compliance rule enforcement?
• RQ2 What role can large language models (LLMs) and NLP play in extracting actionable insights from unstructured data sources like emails and sensor logs for process modeling?
• RQ3 How can emerging AI techniques such as transfer learning and federated learning improve the reliability and scalability of data-driven BPM systems?
• RQ4 What frameworks and metrics are needed to ensure high-quality, trustworthy, and transparent process data in data-driven BPM?
• RQ5 How can process-aware automation be advanced beyond task-level automation by embedding AI-driven decision support and real-time monitoring?

Key findings

• The 'predict and comply' pipeline enables more flexible, maintainable, and transparent compliance monitoring by predicting compliance states before runtime, improving system responsiveness and adaptability.
• Leveraging LLMs and NLP on unstructured data sources such as emails and videos significantly enhances process discovery and monitoring capabilities, enabling richer context-aware process models.
• Predictive business process monitoring, when extended to include compliance prediction, allows for proactive risk detection and improved decision-making with reduced human bias.
• The integration of advanced AI techniques like neuro-symbolic AI and federated learning can enhance model robustness, privacy, and generalization in data-driven BPM applications.
• Measuring and improving process data quality is essential for reliable automation and trustworthy AI-driven recommendations in BPM systems.
• The proposed research agenda identifies five high-impact research areas—data foundations, AI integration, compliance monitoring, data quality, and automation—offering a structured pathway for future innovation in BPM.