Managers anticipate more AI agents than humans in a year

A majority of executives anticipate supervising more AI agents than employees within a year, while fearing they will lose control over their actions. Meanwhile, software stacks are growing rather than simplifying, and performance evaluation is becoming more opaque. A Nielsen survey conducted among 1,700 managers across seven countries for Monday.com paints a picture where governance takes precedence over mere productivity.

AI Agents Disrupt Supervision and Pose Liability Risks

58% of managers expect to oversee more AI agents than employees in the coming year. In this context, 46% fear losing control over the actions of these systems, and 35% worry about being held accountable for their decisions, as language models remain largely unpredictable. Managers do not want fully autonomous agents; they want humans to be involved and consulted by AI on crucial tasks and decisions. Specifically, it is necessary to decide which missions to assign to agents, verify their outputs, and intervene in case of deviations, a supervision that adds a layer of invisible work.

The Proliferation of Tools Undermines Rationalization Hopes

According to 57% of managers, the number of available tools has increased compared to the previous period, while about 51% believe their software has become more interconnected and integrated since the introduction of AI. Only 29% report having a centralized and fully optimized AI infrastructure. Companies thought AI would help reduce their software stacks; instead, they are facing more applications, integrations, and unapproved AI adopted outside validated systems. By the end of 2025, managers aimed for consolidation, but the number of tools has increased: over 50% say they are using more tools than a year ago, and the trend continues.

Evaluating with AI: Desired Efficiency, Confirmed Bias Risks

46% of managers rely on AI for decision-making, and 39% of executives use it to draft performance evaluations. 78% have modified their evaluation grids to emphasize prompt mastery, the ability to delegate to AI, and the experimentation of new working methods. However, using AI to assess individuals can introduce biases: research from the Luxembourg Institute of Socio-Economic Research shows that, without objective standards and during independent evaluations, LLMs score similarly to humans, and a paper published in early 2025 points out biases against Black candidates, varying by group. Companies have yet to figure out how to manage these issues, as over 50% of executives report finding it more difficult to judge work since the introduction of AI, and nearly 50% spend more time controlling and guiding deliverables, often received already very "polished."

Strong Appetite for AI, Hampered by Organizational Adaptation Gaps

94% of managers express enthusiasm, and many believe they know where AI creates the most value. At Monday.com, Michal Miller Levi compares this enthusiasm to that of the early 2000s around the Internet, while emphasizing that AI is advancing faster than internal processes. While the tool benefits managers individually, their role in organizational steering suffers from a lack of adaptation. Companies are deploying AI solutions to gain productivity, but on-the-ground difficulties are accumulating, as revealed by a Nielsen survey of 1,700 managers in seven countries commissioned by Monday.com. Thus, the adoption of AI proves to be as much a governance issue as a lever for efficiency, and without suitable processes or infrastructures, managers struggle to stay on course.