We use necessary cookies for the living book to work, and we'd also like to use anonymous analytics to improve it. Your reflections are always private and we never analyse them. More details in the cookie policy.
FIG. 13
The transformation funnel
100 pilots enter · 10 survive the real-world impact
Source: McKinsey Global Institute · 90% of AI pilots in large companies don't reach production
The four funnel leaks
why brilliant pilots die when scaling
1Messy data
The pilot was designed with clean, curated data. The real organisation operates on data scattered across legacy systems.
Controlled environments deceive: when the pilot leaves the lab, data quality and availability collapse, and the metrics that looked solid stop holding. The pilot worked; the company wasn't ready.
2Cultural resistance
Employees didn't take part in the design and see it as a threat to their roles.
When the pilot is promoted to standard, the organisation that has to adopt it doesn't feel ownership of the change. The rejection isn't of the technology: it's of the fact that it's imposed without their having understood how it affects them.
3Misaligned incentives
Middle managers don't see how AI fits into the KPIs they're measured by.
If measurement and incentive systems still reward compliance with the old process, no middle manager will absorb the cost of changing it. The pilot dies where the operating plan executes.
4IT saturated, no frame
The technology team doesn't have capacity to support a wide deployment.
The ExOS Portfolio block is missing: no formal mechanism to evaluate which initiatives still contribute and which to cancel. Inertia keeps pilots alive much longer than they should be.
Signs this syndrome affects you
If you recognise more than three, it's not an exception: it's a pattern
List built from dozens of real transformations. They're not independent symptoms: they're the old operating system trying to absorb new technology.
→You have pilots that work, but almost nothing in real production.
→You buy tools before redesigning processes.
→Each department experiments on its own without visible coordination.
→Demos impress, but the P&L doesn't move.
→The teams that led the pilot aren't the same that have to scale it.
→Nobody cancels projects. All keep 'going', none advances.
"The technological gap doesn't close by buying technology; it closes by building the capacity to absorb it.»