Ask any team for a picture of their automation estate and you'll get diagrams: BPMN models, integration maps, an architecture slide from two years ago. All of it describes intention — what the process was designed to do. An estate assessment built on intention alone answers the wrong question. The question isn't "what should happen?" It's "what actually happens, how often, and at what cost?"
The gap is always bigger than anyone expects
When we put runtime data next to the diagrams — instance volumes, path frequencies, task durations, wait states, escalations, reassignments — the same patterns surface in estate after estate:
- Dead processes. Modeled, deployed, maintained — and fired four times last year. Their maintenance cost is pure loss, and their migration cost would be too.
- Shadow paths. The happy path in the diagram carries 30% of traffic; the rest flows through exception branches and manual workarounds the model never acknowledged. That's where the real process lives.
- Hotspot tasks. A handful of human tasks where instances pile up for days — usually re-keying, chasing, or triage. These are your first agent candidates, and no diagram will point to them.
- Value skew. Eighty percent of business volume runs through a handful of processes. The modernization budget should be just as skewed — it almost never is.
Why this changes the verdicts
A wrap-or-rebuild decision made on a diagram is a guess about a fiction. Runtime evidence turns it into arithmetic: the dead process gets retired instead of migrated, the shadow-path process gets rebuilt because its diagram was never true, and the stable high-volume workhorse gets wrapped and left alone. In most assessments, the retire list is the biggest surprise — and the fastest saving.
Where process mining fits
When event logs support it, full process mining reconstructs the real flow from data — loops nobody drew, sequences nobody designed. But even without a mining tool, the BPM platform itself holds most of what you need: BAW's instance and task data alone will collapse more assumptions in a week than a workshop season. The barrier is rarely tooling. It's that nobody asked the runtime.
Evidence first, opinions second
Assessment workshops have their place — context, ownership, politics all matter. But start them after the data. A workshop that opens with "here's what your estate actually did last quarter" is a different conversation from one that opens with a diagram from 2019. This is how we run every estate assessment: runtime first, verdicts on evidence.