TLDR: Organizations struggle to apply AI effectively in end-to-end operational processes due to a lack of structured, process-specific data. This research paper argues that Object-Centric Process Mining (OCPM) is the crucial enabler for generative, predictive, and prescriptive AI in enterprise settings. OCPM provides Process Intelligence (PI) by allowing events to reference multiple objects, creating a unified and flexible view of complex processes. This grounding enables AI to diagnose, predict, and prescribe actions for process improvement, making PI indispensable for successful AI adoption in organizations.
In today’s rapidly evolving technological landscape, Artificial Intelligence (AI) is transforming industries and daily life. However, despite its immense potential, many organizations face significant hurdles when trying to implement AI successfully in their core operational processes. The challenge lies in bridging the gap between sophisticated AI techniques and the often messy, dynamic, and interconnected reality of business operations. A recent research paper titled “No AI Without PI ! Object-Centric Process Mining as the Enabler for Generative, Predictive, and Prescriptive Artificial Intelligence” by Wil M.P. van der Aalst argues that Process Intelligence (PI), particularly through Object-Centric Process Mining (OCPM), is the crucial missing link.
The paper highlights a “smartphone paradox” in the context of AI adoption: just as smartphones optimized the edges of work rather than transforming its core, current AI applications often automate individual tasks but struggle to improve end-to-end organizational processes. This is largely because AI needs to be properly grounded in process-related data, which is often structured and organization-specific, unlike the general data AI models are typically trained on.
Understanding Different Facets of AI
The paper categorizes AI into three main types: Generative AI, Predictive AI, and Prescriptive AI. Generative AI, exemplified by models like ChatGPT, creates new content such as text, images, or code, often guided by a prompt. Its goal is to mimic reality or produce creative outputs. Predictive AI, on the other hand, focuses on forecasting future outcomes by learning from past data. This could involve predicting customer churn or anticipating traffic congestion. Lastly, Prescriptive AI goes a step further by suggesting or enforcing actions to optimize specific goals, often involving mathematical optimization to recommend the best decisions under given constraints.
The Power of Object-Centric Process Mining (OCPM)
Traditional process mining analyzes event logs to understand how work is done, identify bottlenecks, and ensure compliance. However, it often struggles with scattered data, a rigid focus on single “cases,” and oversimplified views when events involve multiple interconnected objects. This is where Object-Centric Process Mining (OCPM) comes in. OCPM starts from Object-Centric Event Data (OCED), allowing events to reference multiple objects of different types – for instance, a payment linked to both an invoice and a customer. This approach provides a unified, consistent dataset, offering a single source of truth that reflects the complexity of real-world processes.
By adopting OCPM, organizations can analyze operational activities from any perspective without needing to re-extract or reshape data. This flexibility is vital for uncovering insights, especially at the intersections of different processes and organizational units, which are often overlooked by traditional methods. OCPM essentially provides the lens through which process-related problems, like bottlenecks or deviations, can be clearly identified and translated into solvable problems for AI.
Connecting AI and Process Intelligence: The Five Opportunities
The research paper outlines five key ways in which OCPM, as the foundation of Process Intelligence, enables and enhances various forms of AI:
First, OCPM provides the necessary context for **Predictive AI**. By identifying process-related problems, OCPM can generate training examples (e.g., instances of bottlenecks with associated delays) that predictive models can learn from to forecast future issues.
Second, it similarly supports **Prescriptive AI**. Once a problem is predicted, OCPM can feed data into prescriptive models or optimization algorithms, allowing them to recommend optimal actions or decisions to improve the process, always respecting predefined goals and constraints.
Third, OCPM facilitates the integration of **Operations Research (OR) techniques**. Many process mining computations already use mathematical optimization, and OCPM can further guide these classical optimization, scheduling, and planning techniques, just as machine learning can guide optimization.
Fourth, **Generative AI** can be used to improve human interaction with process mining software. Users can pose natural language questions about process data or capabilities, and GenAI, especially when augmented with Retrieval-Augmented Generation (RAG), can provide accurate answers based on real-time process mining computations rather than general knowledge.
Finally, GenAI can significantly assist in the **preparation of event data**. Data extraction and transformation are often major bottlenecks in process mining adoption. GenAI, with its ability to generate high-quality code (like SQL), can help bridge the gap between proprietary data formats and the standardized Object-Centric Event Data (OCED) format.
Also Read:
- Unlocking AI’s Potential: A New Approach to Self-Evolving Agents
- Pro2Guard: Ensuring LLM Agent Safety Before Incidents Occur
The Path Forward: AI Needs PI
The core message is clear: for AI to truly transform organizations beyond isolated tasks, it must be grounded in a deep understanding of operational processes and their underlying data. Process Intelligence, powered by Object-Centric Process Mining, provides this essential grounding. By combining these powerful approaches, organizations can move beyond merely automating tasks to fundamentally improving end-to-end processes, leading to reduced waste, fewer delays, and enhanced overall efficiency. When processes work, everything works.
