Companies no longer doubt that AI is capable of working. Rather, they ask why its effect is, at least partly, uneven given the clear increase in worker productivity associated with AI. An interesting contradiction arises from the study. On the one hand, AI has become more accessible to workers very quickly (the number has gone up to about 60%). But on the other hand, the actual usage of AI in daily work is still lower than that level, showing that there is a big “activation gap” where the tools are present but the change is not. This gap is not only a matter of human behavior but, at the same time, is a sign of deep structural inertia.
Most companies are still stuck in what could be called the “pilot economy,” where lots of experiments are done but very few are scaled up; only a quarter of them have really made AI a part of their operation. More than half think that they will do so very soon. At the same time, this positive attitude can hide a mistake that keeps happening. The execution failure that transfers from the controlled environment to the real-world introduction in fact brings in integration complexity and governance responsibilities.
Human Resource Issue Vs Design Issue
The unpredicted edge cases that many companies systematically underestimate. What is more, the study highlights the fact that the adoption of AI is being unnecessarily considered a human resources issue when it is actually a design issue. Although firms are making large investments in AI literacy, nevertheless, 84% of them have not redesigned jobs or workflows with AI. In other words, they are practically just adding new technology to old operating models. Therefore, the AI-generated improvements in cost and productivity are very much visible (these are efficiency gains that companies mostly and widely achieve).
AI still fails to become the driver of strategic reinvention. This is evident as only a small proportion of companies are changing their business models or creating new ways of generating value. This difference is steadily helping to increase the competitive gap between the “optimizers” and the “reinventors.”
In parallel, the swift appearance of agentic AI brings a second-level risk. The capability is improving faster than the capacity to manage it. Since only 21% of organizations have a well-developed framework for the governance of autonomous systems, the probability of large-scale systemic operational and reputational failures is increasing.
Conclusion
The political shift toward sovereign AI means that technological choices are becoming more and more restricted by the locality of data, different regulatory regimes, and national strategies rather than only performance factors. This is effectively causing the fragmentation of the global AI ecosystem.
Physical AI is revealing the next major challenge, but its longer time to market is actually a pointer to a very important point: as AI is getting closer to the physical world, it is becoming more and more limited by the need for capital, safety, and regulatory constraints. The three together speak volumes about the fact that the real blockage in AI transformation is not creative thinking but the ability of organizations to adapt.
Companies that treat AI as merely another tool will remain trapped in incremental efficiency gains, while those that redesign their structures, governance, and value creation around AI will define the next wave of competitive advantage.
