Companies are handing out the AI tools to their employees to improve productivity, but these tools seem to have brought new problems for their managers as well.
According to a survey conducted by Nielsen ( Nielsen ) on 1,700 directors and executives in seven countries, commissioned by Monday.com, more than half of the executives stated that after adopting AI, it has become more difficult for them to evaluate their employees' work. The reason is that a large portion of the employees' output is now polished and directly sent to managers' inboxes, making it hard to determine how much of it comes from the employees themselves and how much is from AI. As a result, nearly half of the managers reported spending more time reviewing and guiding their teams' work than before.
This does not mean that managers are dissatisfied with AI. On the contrary, 94% of the managers surveyed expressed excitement about using AI to improve their work, and many of them are confident about where to apply it to achieve the greatest value. Monday.com, the head of market research and customer insights, Michal Miller Levi, compared this enthusiasm to the early days of the Internet in the 2000s, saying that people see a similar "magic" in AI.
But she told TechCrunch, the problem is that the development speed of AI has already surpassed the adaptability of the company. “AI is very effective for [managers] individually, but at the same time they also hold different managerial roles. They also need to drive the entire organization forward... Their dilemma is that the pace of AI advancement is far faster than the rate of change in organizational processes,” said Levi.
Managers are also rethinking how to evaluate teams: 78% of respondents stated that after adopting AI, they changed their approach to evaluating employees, placing more emphasis on employees' ability to create prompts, determining which tasks can be assigned to AI, and trying out new ways of working.
But when managers cannot easily determine how much of a task is generated by AI, how can employees be assessed fairly? What if managers themselves become increasingly dependent on AI to make such judgments? Surveys show that 46% of managers use AI to assist them in making decisions, and 39% of executives use AI to draft performance evaluations.
This approach may introduce more bias into the evaluation process than usual, as LLM is essentially a predictive machine that tends to replicate the human patterns present in the training data. Earlier this year, research from the Luxembourg Institute of Social Economics found that in the absence of clear and objective criteria, and when individuals are scored individually, LLM often gives similar scores in a manner similar to humans. Another research paper from early 2025 discovered that the AI model may be biased against black male candidates and "shows significant deviations, with the direction and magnitude of these deviations varying among different social groups."
When I asked Levi, whether using AI to evaluate employee performance would introduce external biases into the process, she stated that companies are still figuring out how to handle such issues.
AI is also changing the day-to-day work of managers. Someone has to decide which tasks to delegate to AI, check their output, and intervene in case of errors. The report suggests that this kind of supervision may constitute a new form of "invisible labor."
The software itself has also become more difficult to manage. 57% of managers stated that their organizations are using more tools than before; approximately 51% said that after adopting AI, the interdependence between these tools has increased and they have become more centralized. Only 29% believe that their AI infrastructure is fully optimized and centralized.
Levi said that companies originally thought AI would help to reduce the software stack. However, what they faced was more applications, more integrations, and the "shadow AI" phenomenon caused by employees using tools outside of the systems approved by the company.
“The feeling of managers at the end of 2025 was that they would integrate tools and actually reduce the number of software applications. But in reality, the number of tools they used increased,” said Levi. “More than 50% of people indicated that they are using more tools now than last year, and this trend has not stopped…”
As enterprises deploy AI agents that can act autonomously, this need for control seems to become increasingly urgent. Surveys find that 58% of managers expect that within the next year, the number of AI agents they will need to supervise will exceed the number of human employees; 46% are concerned about losing control over the behavior of these agents.
People are also very concerned about having to take responsibility for the decisions made regarding AI (35% of respondents expressed this), as LLM remains largely unpredictable. "They don't want their agents to be completely autonomous," said Levi, and explained that managers hope that in key tasks and decision-making, AI can involve people and seek their opinions.
Essentially, the findings of this report indicate that AI has become more than just an opportunity to enhance productivity; it has also turned into a governance issue. Without appropriate processes and infrastructure in place to address it, managers seem to be left scrambling to maintain operations.
