Process Mining has been the “hot thing” for a few years, but momentum has stalled as organizations work on rolling it out. Gartner classified it as "early mainstream," yet only about 20% of people in talks confirm actually running projects.
In this episode, we sit down with Lukas Pfahlsberger (CEO) and Jan Mendling (Humboldt University Professor), co-founders of Noreja. They explain how their tool uses AI, knowledge graphs, and context like documents to shift away from traditional event logs and uncover root causes.
The Evolution of Process Mining: Started in the 1990s and expanded around 2000 via Wil van der Aalst, originally focusing on simple, linear processes with single objects.
The Complexity Gap: Classical techniques struggle with complex enterprise workflows (like order-to-cash) where objects split, converge, or diverge across systems.
Moving Beyond Correlation: Causal process mining uses a priori knowledge to validate actual triggers rather than assuming correlation equals causation.
Multi-Source Causal Modeling: Noreja builds models using database schemata (keys/cardinalities), AI/LLMs to validate logical steps, and domain knowledge.
Skipping the Flat-File Headache: Direct connection to source systems avoids discarding rich database semantics and saves time spent on manual ETL data pipelines.
Streamlined Onboarding: Connecting via database/API allows users to select core tables and automatically generate an initial entity graph.
Perspectives Over Rigid Views: Users define custom "perspectives" by choosing specific starting objects, letting AI pinpoint root causes within scoped views.
Lowering Barriers to Entry: Companies can immediately spot where data quality issues reside and measure their exact operational impact.
AI Context & Token Efficiency: Graph databases structure processes cleanly for LLMs, saving token costs and avoiding messy "spaghetti diagrams."
Maintaining Dynamic Context: Connecting SOPs and procedural versions lets organizations verify compliance against evolving business rules.
Process Frontier Agents: Autonomous background agents manage context by scanning repositories, archiving outdated files, and organizing knowledge.
Cost-Efficient Local LLMs: Using local open-weight models (like Gemma) during off-hours reduces API costs while handling heavy context tasks.
Compressing the BPM Lifecycle: Combining AI insights and rapid artifact generation aims to shrink 12-month BPM cycles into 2-to-3-day workshops.
Proving ROI: SAP research shows an average 1:20 return on process improvement projects, which accelerates as decision velocity improves.
Next Steps & Resources: Explore Noreja’s whitepapers, blog, or demo environment on their website or reach out via LinkedIn.
