Process AI: Mining and Improving Workflows
Process AI
Process AI tests how to use event data and AI-assisted analysis to improve workflows. The focus is process mining, task mining, process discovery, conformance checking, and automation opportunity assessment. You need to know what evidence each technique produces and how to translate it into a Pega case design change.
Core rules
- Process mining uses case-centric event logs with case ID, activity, and timestamp; it discovers the as-is process and deviations.
- Task mining captures user-level actions and is useful for repetitive desktop tasks, not for end-to-end case flow alone.
- Conformance checking compares observed behavior with a target model; enhancement identifies redesign opportunities.
- Event logs need clean timestamps, consistent activity names, and complete case coverage.
- Automation candidates should be high-frequency, rule-based, low-exception steps; validate with cycle time and rework.
Worked example
An insurance case log shows 40% of cases return from Underwriting to Documents. Reasoning: rework is a stronger signal than average duration. Calculate rework rate and loop count; segment by channel. The team finds missing document checks at intake. They add a validation step before underwriting rather than automating the underwriting approval.
Traps
- Confusing process mining with task mining; one uses case event logs, the other captures user actions.
- Using timestamps at day granularity and concluding a step is fast.
- Mining only happy-path cases and missing exceptions.
- Treating every bottleneck as an automation candidate; some need policy or data fixes.