Why pilots stall: the system nobody changed.
You ran the pilots. The demos worked. People were impressed. Then the quarter closed and nothing had moved. The pilot was never the hard part.
A pilot proves one thing: the technology works. That is worth knowing, and it is also the easiest of the three things that have to change. The other two are where transformation actually lives, and they are the two that most programmes never touch.
Three systems, and the one nobody owns
Every AI transformation moves three systems at once: the people who do the work, the technology they use, and the business around both. Change one and leave the other two behind, and the product is zero. That is the multiplication. It does not add up. It multiplies, and anything multiplied by zero is zero.
Technology gets changed, because that is what the pilot was. People sometimes get changed, through training and a few internal champions. The business system almost never gets changed, because no single person owns it. The AI lead does not set pricing. The workflow owner does not set incentives. So the one system that decides whether any of it pays off is the one nobody is holding.
A time saving the model could not use
At one accountancy firm we worked with, AI saved real time on recurring work. People used it, and kept using it. On paper, a success. But the firm still billed by the hour. Every hour AI saved was an hour the firm could no longer invoice. Adoption was working exactly as intended, and the commercial model quietly turned that gain into a loss.
Adoption alone could not fix that. The model had to move too.
No amount of extra training would have solved it. The constraint was not in the people or the tool. It was in the business system, and until that moved, better adoption only made the gap wider. The gain stayed on paper. That is the multiplication doing its work.
Usage is not change
This is why usage dashboards mislead. They count logins. They tell you a tool was opened. They say nothing about whether a workflow got better, whether an outcome moved, or whether the business can keep it going. A team can be fully adopted and completely stalled at the same time.
The question is whether the work changed, and whether the change survives you leaving.
What changing the system looks like
Start from the constraint. Find the one system holding the value down. Redesign the work with the teams who do it. Move the business conditions that quietly cancel the gain: pricing, incentives, governance. Then transfer ownership, so the capability stays when the outside help leaves.
That last part is the one we intend to measure directly: we call it the Ownership Metric. The share of AI work your team ships without us is the number that tells you the transformation is real. A pilot proves the technology. The transformation is the other two systems.
Read the full breakdown of the Ownership Metric →
← All field notes