What the study shows in retrospect
The picture the study paints for the past three years clearly contradicts the public AI narrative. Roughly 70 percent of surveyed companies in the US, UK, Germany, and Australia now actively use AI - a high adoption rate. Yet nine in ten surveyed executives report no measurable effect on employment or productivity at their own company. In this study, adoption and measurable impact clearly diverge: having a tool in use, per these findings, doesn't automatically mean it shows up in the company's bottom line.
That aligns with a pattern this series has already observed elsewhere - for instance in the Bitkom study, per which a third of German companies report AI costs higher than expected. Together, both studies paint a more consistent picture than the single headline 'AI is changing everything': the costs often arrive faster and more visibly than the benefit.
The more interesting part: the expectations gap
For the next three years, the study paints a different picture. Executives on average expect a 1.4 percent productivity gain, combined with a 0.7 percent decline in employment - net output growth of roughly 0.8 percent. That forecast is markedly more optimistic than the actually observed impact of the past three years.
A second finding is more notable, though: employees at the same companies were also surveyed - and expect exactly the opposite employment trend for the same period, namely a 0.5 percent increase rather than a decline. The study explicitly calls this an 'expectations gap' between employers and employees. That gap matters regardless of which of the two forecasts eventually proves correct: it shows that leadership and staff at many companies are working from fundamentally different assumptions about their own AI strategy - a communication gap with consequences of its own, independent of the actual economic outcome.
Why the impact hasn't arrived yet: an organizational question, not a technical one
A companion study from the same research group, titled 'Mind the Gap: AI Adoption in Europe and the U.S.', offers a concrete explanation for why the impact becomes visible faster in some countries and companies than others. Per the study, the decisive difference doesn't lie in the technology deployed, but in organizational support: whether a company actively encourages its employees to use AI, provides them with suitable tools, and trains them in it - all three elements together, not just one of them.
The numbers are stark: in the US, 42 percent of workers receive all three forms of support simultaneously - encouragement, tools, and training. In France and Italy, that share is only 16 to 17 percent. Conversely, 69 to 70 percent of workers in France and Italy receive none of these three forms of support, versus only 44 percent in the US. Per the study, this organizational gap explains a substantial part of the difference between countries with faster versus slower measurable AI impact - not the underlying availability of the technology.
What this means for your own AI strategy
For a company in our audience, these two studies suggest two separate but connected lessons. First: if no measurable AI effect has shown up in your own company yet, per this study, that's not an outlier but the normal case at nine out of ten companies - a reason to calibrate your own expectations, rather than prematurely concluding a deployment has failed, or conversely promising exaggerated short-term results.
Second, and this is the more actionable point: per the study, whether a company ends up among those that actually achieve impact depends more on organizational support than on the tool deployed itself. A company that rolls out AI tools without actively encouraging their use and without offering training structurally sits in the group less likely to achieve measurable effects - regardless of how capable the deployed model is.
On the expectations gap between leadership and staff: if executives internally assume future job cuts from AI while staff expect the opposite, a communication gap emerges that can have negative effects independent of the actual economic development - for instance on trust, willingness to change, and acceptance of new tools in daily work. Addressing that gap transparently is a leadership step in its own right, regardless of which of the two forecasts later turns out to be correct.
What this means in practice
- Calibrate your own expectations for AI investments against this data: per the study, the absence of short-term, measurable effects is the normal case at nine out of ten companies, not a sign of a failed project.
- Check whether your own company actually provides all three support elements together - active encouragement to use AI, suitable tools, and targeted training - rather than relying on simply making a tool available.
- Actively address your own expectations gap between leadership and staff rather than leaving it unspoken - for instance through open communication about what employment and productivity development your company actually expects over the coming years, and why.
- Don't measure AI success solely by adoption rates (how many employees use the tool), but by actually measured impact - the study clearly shows the two metrics can diverge widely.
The real value of this analysis isn't reassurance or a warning against AI investment, but calibration: the study provides a solid, data-backed basis for treating the absence of short-term effects as normal - and for locating the decisive lever for actual impact in your own organization rather than in the AI tool you've chosen.