The Pilot Trap
Twelve pilots that succeeded. The question is never whether it worked — it is what the rollout will not have: the volunteers, the vendor engineer, the director watching weekly.
1. A workflow tool piloted with the two teams that asked to be in the pilot. Adoption reached 94%.
2. An AI drafting assistant piloted across a randomly selected 10% of the customer service team, with the same training and support everyone would get. Handling time fell 18%.
3. A new expenses system piloted with a vendor engineer embedded with the team for six weeks. Satisfaction was high.
4. A knowledge base piloted in one department. “It went well and people liked it.”
5. A scheduling tool piloted across three sites chosen because they represent the range of site sizes. Two project staff supported it for the first fortnight only, as planned for rollout.
6. A CRM piloted with the sales team, whose director championed the business case and reviewed usage weekly.
7. A document automation pilot run with the legal team's most experienced drafters. Error rates were negligible.
8. A field service app piloted with engineers in one region over four weeks. The measure was job completion time against the same period last year, agreed in advance.
9. A procurement portal piloted with suppliers who agreed to take part. Onboarding time halved.
10. A chatbot piloted on the internal helpdesk, with a human monitoring every conversation and stepping in silently when it went wrong.
11. A time-recording system piloted for one month across two departments. “Users found it intuitive.” No completion or accuracy data was collected.
12. An asset management tool piloted with the team that maintains the current spreadsheet, who volunteered and built the configuration themselves.
Educational. Pilot design varies; what transfers is asking who was in it, what help they had, and what was written down as success before it started — then asking what the rollout will not have.