Although access to AI tools among employees has increased by 50% over the past year, businesses are struggling with a growing gap between the availability of technology and its actual activation. The 2026 edition of the Deloitte report "State of AI in the Enterprise"shows that most organizations are still struggling with the transition from small-scale pilots to full production.
For the study, Deloitte surveyed 3,235 executives worldwide about their AI initiatives. What did they find? While three-quarters of companies consider AI essential to their strategy, 84% of organizations have not yet redesigned the nature of work or roles within the company.
As a result, the technology is often deployed as an additional layer on top of existing, unchanged work processes, which limits the expected productivity gains.
This stagnation manifests itself in what researchers call the "Pilot Trap." Only 25% of respondents indicate that more than 40% of their AI experiments have actually been integrated into daily operations.
The barriers to scaling up are often not technological in nature, but lie in a lack of robust data infrastructure and the scarcity of talent capable of working with these systems.
"Organizations are now moving from experimentation to large-scale deployment of AI and increasingly integrating it into their core activities," says Nitin Mittal (global AI lead at Deloitte). "AI must be consciously integrated into work processes and enable better collaboration between human and machine intelligence in order to realize its full value."
Whereas attention was previously focused primarily on software applications, Deloitte now signals the definitive breakthrough of 'Physical AI'. This concerns the integration of intelligence into robots, drones, and sensors that make physical decisions in the outside world.
Currently, 58% of companies already use this technology, a share that is expected to rise to 80% within two years.
The advantages of this trend are considerable; it enables companies to directly optimize their physical operations, such as logistics or security monitoring, by linking data to hardware. However, the threshold is high: while software pilots are relatively inexpensive, Physical AI requires investments in the millions due to the necessary hardware and modifications to the physical infrastructure.
A notable new development in the report is the growing focus on "Sovereign AI." No less than 77% of companies now consider the geographical origin of AI solutions in their choice, driven by concerns about geopolitical dependence and data sovereignty.
Meanwhile, the market is preparing for the emergence of "Agentic AI": autonomous systems that can perform actions independently. Although 74% of companies want to implement this within two years, only 21% currently have a mature governance model to manage these agents.
Jorg Schalekamp (AI lead at Deloitte Netherlands) emphasizes that competitive advantage in the near future will come from this human adoption. "By strengthening both the capabilities of their talent and the AI tools, teams can apply new working methods."
According to the report, the biggest challenge for 2026 is therefore not acquiring more computing power, but strategically reinventing how organizations achieve their goals.
