While corporations continue to pour billions of dollars into artificial intelligence tools with the primary promise of making their workforces drastically more productive, these same technologies are quietly spawning an entirely new generation of operational and administrative headaches for middle and senior management.
According to a comprehensive new survey commissioned by work-management platform Monday.com and conducted by Nielsen, the integration of generative AI into daily workflows has fundamentally upended traditional structures of oversight. The poll, which surveyed 1,700 directors and senior managers across seven countries, reveals that more than half of senior leaders now find it noticeably harder to evaluate their employees’ work accurately.
The root of this evaluation crisis lies in the polished nature of modern AI output. Because employees are increasingly relying on advanced language models to draft reports, write code, formulate strategies, and polish communications, much of the work arriving in a manager’s inbox is pristine and professional on the surface. However, this high sheen makes it exceptionally difficult for leadership to discern the true provenance of the labor—blurring the line between an employee’s original analytical thinking and the algorithmic generation of an AI tool. Consequently, nearly half of all managers report spending significantly more time carefully reviewing, validating, and guiding their teams’ output than they did prior to the widespread adoption of these technologies.
Yet, this administrative friction does not mean that leadership has soured on the technology. Quite the opposite sentiment prevails across corporate offices. An overwhelming 94% of the managers polled expressed genuine excitement about the potential to improve work processes using AI, and many exhibit a high degree of confidence regarding where and how to deploy these tools to extract maximum business value.
Michal Miller Levi, head of market research and customer insights at Monday.com, draws a direct parallel between current corporate enthusiasm and the exuberant early days of the commercial internet back in the 2000s, noting that professionals today still perceive a distinct, transformative kind of "magic" in artificial intelligence.
However, the fundamental disconnect, Levi explained in an interview, is that the rapid evolution of artificial intelligence is outpacing the ability of organizations to adapt their internal structures and operational processes. AI functions remarkably well for individual contributors and managers alike in isolated tasks, but leaders must simultaneously shoulder a broader organizational responsibility. Their current professional struggle stems directly from the stark reality that AI development is moving at a velocity far exceeding traditional corporate adaptation cycles.
This shifting dynamic is also forcing a profound reevaluation of how performance is judged. A striking 78% of survey respondents admitted that they have actively changed their evaluation criteria for employees following the adoption of AI. Traditional metrics of output volume and baseline execution are taking a back seat, replaced by a heavier emphasis on an employee’s proficiency in prompt engineering, their strategic judgment in deciding which tasks to delegate to algorithms, and their willingness to experiment with novel, AI-driven ways of working.
This evolution in performance evaluation, however, introduces a thicket of ethical and practical dilemmas. How fairly can an employee be judged when their supervisor can no longer accurately gauge the human versus machine composition of their deliverables? The problem is further compounded by the fact that managers themselves are increasingly turning to AI to solve their administrative burdens. The survey revealed that 46% of managers now use AI systems to help inform their professional decisions, while 39% of leaders utilize artificial intelligence to help draft employee performance reviews.
This widespread reliance on automated tools for human evaluation carries significant institutional risks, chief among them the potential for amplifying systemic bias. Large language models are, at their foundational level, sophisticated prediction engines designed to replicate patterns found within their vast training data. Without rigid oversight, they tend to mirror human prejudices.
Research published earlier this year by the Luxembourg Institute of Socio-Economic Research highlighted that in the absence of clear, objective evaluation standards and when individuals are rated independently, large language models tend to replicate traditional human evaluation biases. Furthermore, a research paper published in early 2025 demonstrated that various AI models can exhibit significant biases against specific demographics, such as Black male candidates, with biases fluctuating in direction and magnitude across different social groups.
When asked whether leveraging AI to weigh in on human performance risks injecting outside prejudice into the corporate review pipeline, Levi acknowledged that businesses have largely failed to figure out how to navigate and resolve these complex ethical hurdles.
Beyond evaluations, artificial intelligence is radically reshaping the daily responsibilities of a modern manager. Someone within the corporate hierarchy must constantly decide which granular tasks to hand over to autonomous AI agents, meticulously check the resulting output, and manually step in whenever a workflow goes awry. The Monday.com report suggests that this ongoing oversight amounts to a massive, unrecognized layer of "invisible labor" that silently consumes executive bandwidth.
Compounding the problem is the increasing complexity of the software ecosystem itself. Fifty-seven percent of managers reported that their organizations now have a significantly higher volume of disparate software tools in circulation than they did prior to the AI boom. Furthermore, about 51% noted that their tech stacks have grown more interdependent and clustered since adopting AI. In stark contrast to these figures, only 29% of respondents described their current AI infrastructure as fully optimized and centralized.
Levi noted that businesses initially entered the AI era anticipating that these technologies would help shrink and streamline their software stacks. Instead, organizations are grappling with a proliferation of new applications, complex integrations, and widespread "shadow AI" usage—a phenomenon where employees independently adopt unauthorized external tools outside of approved company systems.
Managers closed out the previous year with the firm expectation that they would be able to consolidate their software ecosystem and reduce the sheer volume of applications they rely on, Levi observed. Instead, the exact opposite occurred. More than 50% of managers indicate they are utilizing a greater number of tools now than they were just a year ago, and that upward trajectory shows no signs of slowing down.
This urgent need for governance and control is becoming increasingly critical as companies deploy autonomous AI agents capable of executing complex workflows independently. The survey found that 58% of managers expect to oversee a larger volume of AI agents than human employees within the next year alone. At the same time, 46% admit they worry about losing operational control over what those autonomous agents actually do in practice.
Accompanying this loss of control is a palpable fear of personal liability. Roughly 35% of respondents expressed deep anxiety over being held professionally responsible for errors, hallucinations, or poor decisions made by autonomous AI systems, given that large language models remain inherently unpredictable. Leaders do not want their digital agents operating in a vacuum, Levi explained, emphasizing that managers want human oversight systematically built into critical tasks and high-stakes decisions.
At its core, the findings of the report underscore that the widespread integration of artificial intelligence is no longer solely an opportunity to unlock unprecedented levels of productivity. Rather, it has evolved into a critical governance challenge. Without the implementation of robust internal processes, clear standards, and optimized infrastructure to manage the fallout, modern managers are increasingly left scrambling just to keep their organizations running smoothly.
