Insights/AI in Practice
AI IN PRACTICE

Shadow workflows are quietly holding your adoption numbers up

21 August 20268 min readAIP-3, AIP-6L5

Adoption reporting after go-live measures use of the new system. It does not measure retirement of the old one. Those are different questions and only the second one tells you whether anything changed.

The spreadsheet did not die

People are conscientious. When a new workflow is slower, or less trusted, or simply unfamiliar under deadline, the rational response is to do both: enter the record in the new system and keep the real work in the old one. Usage goes up. Nothing has changed.

This is not resistance and it is not a training problem. It is a design signal. A parallel process persists because the new path has a gap in it, and the people running it found the gap first.

L5Workflow and Boundary Readiness asks whether the work was actually redesigned, not whether a system was deployed

What to instrument instead

AIP-3 measures whether the prior path was retired, which is a different question from whether the new one is used. AIP-6 instruments the behavioural signal that says whether the new path is carrying real decisions or ceremonial ones.

Together they answer the question adoption dashboards cannot: is the work happening here now, or is it happening somewhere we stopped looking.

Usage is an activity measure. Retirement is an evidence measure. Only one of them is hard to fake.

The practical test is unglamorous. Ask what would break if the old system were switched off on Friday. If nobody can answer, it is still carrying load.

This article draws on AIP-3, AIP-6See how these instruments sit in the framework

Start with the free diagnostic

Ten questions, an indicative readiness band, and one prescribed next step. Returned on screen, no call required.

See what it measures