Key takeaways
- Managers report difficulty assessing how much of their teams' work comes from AI versus human effort, complicating performance evaluation.
- Despite expectations that AI would streamline software, 57% of managers now oversee more tools than before, with 51% dealing with tighter integrations.
- Within a year, 58% of managers expect to supervise more AI agents than human employees, raising concerns about autonomy and accountability.
- Relying on AI to draft performance reviews poses risks of amplifying bias, as LLMs tend to replicate patterns from their training data.
Companies worldwide are installing artificial intelligence tools to boost team productivity. But a Nielsen survey of 1,700 directors and senior managers across seven countries, commissioned by monday.com, reveals an emerging problem: the tools are making managers’ jobs significantly harder. More than half of these senior leaders now find it harder to evaluate their employees’ work, and nearly half spend more time reviewing and guiding their teams’ output than they did before. The reason is straightforward: employees’ work arrives polished and complete, making it nearly impossible to tell how much came from the person and how much from the AI.
The paradox is not lost on managers themselves. Ninety-four percent said they are excited to use AI to improve work, and many expressed confidence about where to deploy the technology for maximum impact. Michal Miller Levi, head of market research and customer insights at monday.com, told TechCrunch that the enthusiasm mirrors the early days of internet adoption in the 2000s, when people saw similar “magic” in the emerging technology. The challenge, however, is pace: AI is advancing faster than organizations can adapt their processes and governance structures to handle it.
Rethinking How Work Gets Evaluated
The adoption of AI has fundamentally changed what managers look for when assessing employee performance. According to the Nielsen survey, 78% of respondents altered their evaluation criteria after implementing AI tools. Managers are now placing greater weight on how well staff members craft effective prompts, decide which tasks to delegate to AI, and experiment with new workflows rather than on traditional measures of output quality or volume.
The Visibility Problem
The challenge is that polished work obscures the human contribution. Managers cannot easily determine whether an employee did the work, collaborated with AI, or let AI handle most of it. This opacity makes performance differentiation difficult at precisely the moment when it should matter most. The intermediate steps and working processes, once clear indicators of competence and problem-solving approaches, have become invisible, leaving managers uncertain about actual individual capability.
The Risk of AI-Assisted Reviews
Compounding the evaluation challenge is managers’ own increasing reliance on AI. The survey found that 46% of managers use AI to help inform their personnel decisions, and 39% of leaders are using AI to draft performance reviews. This practice introduces fresh bias risks into a process already prone to human error. Researchers from the Luxembourg Institute of Socio-Economic Research found that when clear objective standards are absent, LLMs rate performance similarly to humans, which means they repeat the same patterns and prejudices present in their training data. A research paper published in early 2025 identified that AI models demonstrate significant bias against certain demographic groups, with different biases appearing across different populations. When AI systems help generate the language and reasoning behind performance reviews, these algorithmic biases become embedded in official employment decisions with lasting consequences.
The Tool Stack Explosion
Companies began adopting AI expecting to streamline their software environments. Instead, the opposite has occurred. Fifty-seven percent of managers say their organizations now have more tools than before adopting AI. About 51% report that their tools have become more interdependent and clustered in complexity. Only 29% describe their AI infrastructure as fully optimized and centralized.
The disparity reflects a deeper organizational reality: managers expected AI to reduce software sprawl, but the technology has instead enabled proliferation. Employees are adopting tools outside company-approved systems, creating shadow AI usage. Meanwhile, integrations between new AI platforms and existing software have become more intricate, making any single change risky and any comprehensive audit difficult. Levi noted that this outcome contradicts management’s stated intent, saying that managers “completed 2025 with the feeling that they will consolidate tools” but instead “the number of tools that they’re using increased” and “it’s not stopping.” More than half of respondents are using more tools now than they did last year.
The Shift Toward Autonomous Agents
The survey reveals emerging expectations about AI agents taking on decision-making roles. Fifty-eight percent of managers expect to oversee more AI agents than human employees within the next year. This transition from AI as a productivity assistant to AI as a semi-autonomous agent represents a significant shift in organizational structure and control. Unlike a spreadsheet or a design tool that requires human input at each step, autonomous agents can initiate actions, make decisions, and implement changes with minimal oversight.
Control and Accountability Concerns
The prospect of agent autonomy is generating anxiety. Forty-six percent of managers worry about losing control over what agents do, and 35% fear being held responsible for decisions made entirely by AI systems. This concern reflects the unpredictability of current large language models: they can fail in surprising ways, hallucinate facts, or take unintended actions. Managers want to retain some human involvement in crucial decisions, Levi explained, rather than handing full autonomy to agents. The tension between improving efficiency through automation and maintaining organizational control has become acute.
The Invisible Labor of Oversight
As organizations deploy more AI agents, someone has to make critical decisions: which tasks to hand off to AI, whether the output is acceptable, and how to recover when something goes wrong. This supervisory work is becoming a new category of labor that often goes unaccounted for in productivity metrics. It is not the work that gets delivered to clients or published internally; it is the work of managing the systems that supposedly reduce work.
The governance gap is equally significant. AI adoption was supposed to follow organizational processes, but instead it is outpacing organizational ability to manage it. Infrastructure has not kept up with deployment. Policies have not caught up with capability. Training has lagged behind tool releases. Managers are left improvising solutions and creating ad-hoc processes just to keep operations running.
What Managers Actually Want
Despite the challenges, managers are not calling for a slowdown in AI adoption. Rather, they are asking for different kinds of support. They want clearer decision-making frameworks for identifying which tasks belong with humans and which belong with AI. They want better visibility into what AI systems are doing on their behalf. They want accountability mechanisms that do not simply push responsibility downward to individual managers when something fails.
Most fundamentally, managers want AI agents that are responsive to human input rather than fully autonomous. They want guardrails, consultation, and the ability to stop or redirect agents when needed. They want AI to amplify human decision-making rather than replace it, especially in matters that carry organizational or personnel consequences.
Where Adoption Meets Governance
The Nielsen survey illustrates that AI adoption is becoming as much a governance and organizational problem as it is an opportunity for productivity gains. Companies deployed AI tools with relatively little forethought about how managers would assess work, how tool proliferation would be controlled, or how autonomous agents would be overseen. The result is organizational friction: managers are excited about the potential but overwhelmed by the execution, and the processes to handle this new layer of technology are trailing far behind the technology itself.
Source: TechCrunch
Frequently Asked Questions
Why are managers struggling with employee evaluation after adopting AI?
Much of employees' output now arrives highly polished in managers' inboxes, making it difficult to distinguish human contribution from AI-generated content. This visibility gap complicates assessment of individual performance and has led 78% of managers to change their evaluation criteria.
How has AI affected the tools managers need to oversee?
Rather than consolidating software as expected, 57% of managers now have more tools than before. About 51% report their tools have become more interdependent and clustered, with only 29% describing their AI infrastructure as fully optimized and centralized.
What concerns do managers have about autonomous AI agents?
Within a year, 58% of managers expect to oversee more AI agents than human employees. However, 46% worry about losing control over agent actions, and 35% fear being held responsible for AI-driven decisions.
