‘A successful pilot proves it will work everywhere.’

A pilot has delivered excellent results: shorter lead times, satisfied users, a convinced leadership. The decision follows naturally: roll it out to the whole company, reproducing what worked.

But the pilot succeeded under particular conditions: a chosen scope, carefully prepared data, a willing team, often shielded from the rest of the organisation. At scale, these conditions disappear. Other departments, with different practices, come into play, data is more heterogeneous, volumes change, security constraints tighten. A pilot proves that the idea works, not that it will work everywhere.

• Pilots are chosen to succeed.That is legitimate, but their results do not transfer as they are.
• Volunteers are not representative of all teams.Those who come next did not choose the tool.
• The effects on other departments only appear at scale.What the pilot was quietly shifting becomes visible when everyone joins in.

1. 01Before rolling out, list what set the pilot apart from the rest of the company.
2. 02Extend in stages, starting with a scope different from the first.
3. 03At each stage, measure the effects on neighbouring departments, not just the local result.

We help prepare scaling up during the pilot, by looking at what surrounds it.

• Why do our pilots never scale up?→
• Can automating one department unbalance another?→
• Should you know how to stop an AI project?→
• How do you measure the return on investment of an AI project?→