Governance & delivery
Why does a promising digital pilot fail to become normal work?
A pilot often succeeds because the project team quietly supplies the missing operating model.
Published 25 September 2026
List what the pilot borrowed
Record the selected users, clean data, manual fixes, unusual permissions and extra help that made the trial work. Those conditions are legitimate for learning but cannot be assumed at scale.
Examine excluded cases and the days when the specialist was absent. That is where unclear instructions, ownership and recovery routes usually surface.
Build the everyday arrangement
Decide who trains new users, changes a rule, resolves an exception and supports a failure. Test the performance and integrations needed at normal volume, not only in a demonstration.
Expand in a way that preserves learning. A slower rollout with a business owner and a visible support route can be more valuable than multiplying the pilot configuration.
Five volunteers test a workflow while a specialist answers every question. The wider team has less time and many more exceptions.
Before expansion, repeat the test with representative cases and ordinary support arrangements. The question is whether the service works without the specialist in the room.