AI concepts
Process mining (reconstructing processes from system logs)
A technique for reconstructing how business processes actually run, based on system logs (ERP, CRM, BPMS) rather than on documentation or interviews. It shows what really happens, not what is supposed to happen. The gap between the documented picture and the real one is typically 40 to 60 percent.
Primary source: Wil van der Aalst, Eindhoven University of Technology, Celonis State of Process Mining 2025
The term was defined academically by Wil van der Aalst at the Eindhoven University of Technology in 2004. Commercialisation came through Celonis (founded 2011, IPO 2024). By 2025 process mining had become one of the four prerequisites for an effective AI deployment, alongside data quality, governance and C-level sponsorship.
The technical mechanism
Every enterprise system logs events with a timestamp, a case identifier and an activity identifier. Process mining consumes these logs (typically from SAP, Oracle, Dynamics, ServiceNow, Salesforce) and reconstructs the topology of the process: which activities follow which, in what order, with what frequency, how long they take and where errors arise.
The result is measurable in a way that process documentation is not. A BPMN map drawn by an operations manager shows six steps. Process mining over three months of ERP logs shows fourteen, four of which are workaround variants created by people whose specific problem was never addressed by the official procedure.
Why it matters for AI
Deploying AI into a disorderly process does not fix the process, it multiplies its pathologies faster. The McKinsey Global AI Survey of November 2025 found that companies with a measurable AI impact on EBIT are twice as likely to redesign end-to-end business processes before choosing a model. Process mining is the foundation of that redesign, because it shows exactly what needs to be redesigned.
Limitations
Process mining requires data. Unlogged processes (decisions made in meetings, email correspondence, local spreadsheets) are not visible. That is why it is valuable to complement process mining with interviews with key users, as we do in Process mapping.
The Polish context
Among mid-sized Polish corporations, Celonis and similar tools are used by roughly 15 percent of companies with 500-plus employees (our own estimate based on Computerworld Polska surveys, 2025). The rest map their processes through interviews and spreadsheets, which gives the documented picture rather than the real one.