According to the second edition of the **Greenworking x Ipsos bva Observatory**, conducted in May with 1,100 employees and 474 managers, the use of artificial intelligence (AI) in the private sector has accelerated. Seventy-three percent of the participants reported using AI at work, an increase of 10 percentage points compared to the previous year. Among these users, 55% reported using AI regularly or daily, up by 13 percentage points. This growth is largely driven by employees themselves, as 45% of companies have officially deployed AI, a 9 percentage point increase. This suggests that the use of AI without formal approval—known as shadow AI—is widespread. The observatory highlights that shadow AI is the most common practice in the workplace, indicating that many companies are slow to integrate AI officially. Autonomous and informal AI use is more common in organizations with 50 to 99 employees, where 41% are managing AI integration. This increases to 51% in companies with 250 employees or more. The sector of activity also influences the involvement of management. Service companies report an integration rate of 52%, while the education, human health, and social action sectors lag behind at 28%. Not all employees experience the same benefits or challenges when it comes to AI. The observatory found that 75% of users notice an increase in operational speed due to AI. However, this time saved is often quickly absorbed by new tasks. Only 37% of professional users systematically use AI tools approved by their employer, while 54% inform their managers about their AI use. Managers are also heavy users of AI, with 83% reporting usage, including 43% who use tools approved by their employer. According to Greenworking, managers are caught between the need to experiment with AI to stay competitive and the responsibility to enforce compliance rules. The prevalence of shadow AI is attributed to the limitations of current technological integration, as institutional use cases are often limited to basic functions. The study emphasizes the increased workload caused by AI, especially for frequent users. While 52% of users report a decrease in their mental load, the time saved by AI is often offset by the need for more controls and verification of data. The shift in cognitive effort is from direct production to systematic auditing of AI outputs. These added responsibilities place increased pressure on middle management, a group that is expected to shrink by 2030 due to AI, according to a report by BCG released on September 30.