Where Swiss organisations really stand on AI adoption.
Six months. 25 Swiss organisations. 230,000 employees. 12 industries. Designed with Dr. Joël Krapf at Bern University of Applied Sciences. Published in full below, because findings that sit behind a form do not get read.
How we did it
We interviewed and surveyed the people accountable for AI adoption across 25 Swiss organisations spanning 12 industries and 230,000 employees, from 100 person firms to 30,000 person institutions. We asked what had been rolled out, what was measured, and what had actually changed in daily work. Then we compared what leaders reported with what their own numbers showed.
What we found
The gap is not tools, it is transfer
Almost every organisation had deployed something. Very few could point to work that was being done differently as a result. Without structural support, only 10 to 15% of formal training reaches daily behaviour, and self service learning reaches the people who were already curious.
Measurement stops at the licence
75% of organisations measure licences and usage. 6% measure behaviour. That gap explains why adoption programmes report progress while the work stays the same: usage is not capability, and capability is not value.
The blockers are human
Uneven skills, unaddressed fear about roles, governance that arrives late, and training that does not transfer. Where cost cutting was the stated priority, employees disengaged from AI rather than leaning into it.
Governance accelerates adoption
This surprised us. Employees who trust the rules experiment more, not less. Published, applied and revisited guardrails were associated with higher experimentation, because people stop guessing where the line is.
Leadership behaviour is the clearest divider
The variable that most cleanly separated organisations that scaled from organisations that stalled was whether leaders used the tools themselves. Efficiency narratives activate anxiety. Curiosity narratives activate participation.
The distance is widening
Most organisations sit between experimentation and impact. The leaders are compounding, because capability makes the next use case cheaper to find. The laggards are still counting licences.
What follows from this
- Design for transfer, not for events. The weeks between sessions should do as much work as the sessions.
- Measure behaviour before and after. If you cannot say what people do differently, you cannot say the programme worked.
- Publish governance early. It is an accelerator, not a brake.
- Get leaders using the tools before they decide about them.
Want the full report?
The complete write-up, including the maturity model and the measurement instrument, is available as a PDF.