The Context
After stabilizing automated triage and routing (85% accuracy, 1 FTE/month saved), the manufacturer faced a new challenge: scaling automation across a fragmented process landscape.
Documentation described 70 workflows.
Yet tickets told another story—repeated exceptions, cross-team detours, and mismatched categories hiding operational drift.
Leaders suspected process debt.
Data confirmed it.
The Challenge
The company’s automation roadmap depended on reliable process documentation — but official SOPs reflected intent, not reality.
Regional variations, informal workarounds, and undocumented exceptions created silent bottlenecks.
Tickets in Germany took 40% longer to close than identical ones in US.
SAP/AD integrations failed inconsistently.
And the same request type followed three different approval paths, depending on who filed it.
The leadership’s question was simple: “How does work actually flow?” The data held the answer — but buried in 250,000+ unstructured Jira issues.
Our Approach
Arti applied the Knowledge Refinery method to reconstruct actual workflows from raw ticket data:
-
Extracted task sequences, dependencies, and transitions directly from Jira logs
-
Clustered similar patterns using semantic and structural similarity
-
Mapped “shadow” workflows that diverged from official SOPs
-
Visualized gaps between documented and executed processes
The result was a data-driven process map reflecting how work truly happened.
Results
The analysis surfaced far more than undocumented workflows — it revealed how work actually happened.
Arti’s system reconstructed real execution paths, exposing the structural friction buried in daily operations:
-
200+ real workflows uncovered vs. 70 documented
-
Broken approval chains causing geographic delays now visible and corrected
-
SAP/Active Directory integration gaps identified through recurring workaround patterns
-
Reopen and stall patterns traced to missing or misplaced data fields in overloaded Jira forms
-
Workflow-based intake redesign replaced IT-centric forms with adaptive prompts — requesting only the data each workflow truly required
Impact
What emerged wasn’t just cleaner data — it was organizational self-awareness.
Arti turned fragmented Jira history into a single operational map showing how processes evolve, diverge, and fail in the real world.
For the first time, leadership could see exactly:
-
Where approvals stalled
-
How regions diverged
-
Which workflows deserved automation next
The company moved from assumption-driven process design to evidence-based improvement — and built an execution architecture grounded in truth, not documentation.
