Process Mining Pioneer Wil van der Aalst Discusses New Object‑Centric AI‑Driven Process Intelligence

On May 27, Wil van der Aalst, the "father of process mining" and a member of the German Academy of Sciences, gave a keynote at Tsinghua University's Data Intelligence Lecture series, focusing on “Why AI Needs Object‑Centric Process Mining.”

Professor Wang Jianmin, dean of Tsinghua's School of Software, recalled Aalst’s influence since his 2004 book Workflow Management and the 2013 international workflow conference, noting the growing importance of aligning AI with process intelligence.

Aalst observed that over 95% of enterprise AI projects fail because they lack a structured, factual understanding of real business processes. Traditional process modeling relies on manually designed diagrams that often diverge from reality, while classic process mining, though data‑driven, treats a single case as the focal point and cannot handle multi‑object interactions, leading to information loss and chaotic structures.

He introduced the concept of Object‑Centric Process Mining (OCPM), which extends traditional frameworks to link events with multiple objects such as resources, orders, equipment, and materials. OCPM can explicitly represent interactions, concurrency, and dependencies, providing an explainable, implementable view of complex processes and serving as a crucial foundation for deploying AI in enterprise settings, mitigating large‑model hallucinations.

Aalst also explored the frontier of combining generative AI with process mining. He warned against directly applying generic large language models and advocated building domain‑specific foundation models that ingest event data, preserving structural information for interpretability and accuracy. Additionally, he proposed encapsulating process‑mining capabilities as intelligent agents, offering a Copilot‑style natural‑language interface that lowers technical barriers while ensuring that underlying operations are driven by real data rather than purely generated responses.

During the Q&A, Aalst discussed object model definitions, migration costs from traditional process models, and the collaborative boundaries between AI and process mining, sparking an in‑depth academic exchange with students and faculty.