AivexaNewsSearch
AI news for builders and product teamsChecked every hour

AI promises productivity, but for managers, it’s mostly a headache for now

Collected Oct 21, 2026

A survey of 1,700 directors and senior managers across seven countries, conducted by Nielsen for the work-management company Monday.com, found that AI adoption is complicating the manager's job even as enthusiasm for the technology runs high. More than half of the senior leaders polled said they find it harder to evaluate employees' work after adopting AI. The core reason given: output now often arrives already polished, so a manager cannot easily separate what the employee produced from what the model produced.

The result is more supervision, not less. Nearly half of respondents said they spend more time reviewing and guiding their teams' output than they did before. That is a notable inversion of the usual productivity pitch, in which AI is supposed to remove routine work from the queue. Here the review step appears to be expanding.

Enthusiasm, at least, is not the bottleneck. Ninety-four percent of the managers said they were excited about improving work with AI, and many reported confidence about where to apply it for the most value. Michal Miller Levi, head of market research and customer insights at Monday.com, likened the mood to the early internet era of the 2000s, describing a similar sense of magic around the technology. Her read, as she told TechCrunch, is that the difficulty is one of pace: individual managers get real value from AI, but their second job is running the organization, and the tooling is moving faster than the processes built around it.

Evaluation criteria are shifting accordingly. Seventy-eight percent of respondents said they have changed how they assess employees since adopting AI, putting more weight on how well staff write prompts, how well they decide which tasks to hand to AI, and how willing they are to experiment with new ways of working. Those are reasonable signals for a world where delegation to a model is part of the job. They are also harder to measure consistently than a finished deliverable, which is presumably why the review burden is rising.

Managers are also outsourcing part of the judgment itself. Forty-six percent said they use AI to help inform their decisions, and 39% of leaders said they use it to draft performance reviews. The survey poses the obvious tension: how fairly can an employee be judged when the manager cannot tell how much of the work was machine-generated, and when the manager increasingly leans on a model to make the call?

There is research worth weighing here. Work from the Luxembourg Institute of Socio-Economic Research earlier this year found that, absent clear objective standards and when individuals are rated independently, large language models tend to rate performance much as humans do. A separate paper from early 2025 found that AI models can be biased against Black male candidates and show significant biases that vary in direction and magnitude across social groups. Because LLMs are fundamentally prediction engines trained on human-generated text, they inherit the patterns in that data, including the unfair ones. Asking one to weigh in on a person's performance therefore carries a real risk of laundering existing bias into a process that people expect to be principled. Asked whether AI-assisted evaluation could import outside bias, Levi said companies have not yet worked out how to handle the issue.

The manager's day is changing in other ways too. Someone has to choose which tasks go to AI agents, inspect what comes back, and intervene when an agent goes wrong. The report frames this oversight as a new layer of invisible labor, work that consumes time without appearing on any plan. This is a familiar pattern in tool adoption: automation absorbs the visible task and leaves behind a residue of monitoring and exception handling that is easy to underestimate when the business case is written.

The software estate is becoming harder to manage as well. Fifty-seven percent of managers said their organizations now have more tools in circulation than before, and about 51% said their tools have grown more interdependent and clustered since AI adoption. Only 29% described their AI infrastructure as fully optimized and centralized. Levi said businesses had expected AI to shrink their stacks, but instead they are coping with more apps, more integrations, and shadow AI, where employees pick up tools outside the company's approved systems.

Her summary of the year captures the gap between plan and outcome: managers came into the period expecting consolidation, but tool counts went up, with more than half reporting they use more tools now than a year earlier and no sign of the growth stopping.

The governance pressure intensifies as agents gain the ability to act on their own. Fifty-eight percent of managers expect to oversee more AI agents than human employees within the next year, and 46% worry about losing control over what those agents do. Thirty-five percent said they fear being held responsible for decisions made by AI, which is understandable given how unpredictable LLMs remain. Levi said managers do not want their agents fully autonomous; they want a human in the loop, consulted by the AI on critical tasks and decisions. This suggests the near-term market is less about autonomy than about permissioning, approval flows, audit trails and clear ownership of an agent's actions.

Taken together, the findings cast AI adoption as a governance problem as much as a productivity opportunity. The numbers describe organizations adding capability faster than they add process, and managers absorbing the difference in their own calendars.

Why it matters: Product teams building AI features for the enterprise should expect buyers to care as much about oversight as about output, since the survey ties adoption to heavier review loads, larger tool estates and anxiety about accountability for agent actions (inference). Managers, meanwhile, are being handed a second job, judging work whose provenance is unclear, often with AI assistance of uncertain fairness, and that has direct implications for performance-review tooling and how internal AI policies are written. The stated appetite for human-in-the-loop control points to a concrete near-term requirement: designs that make it easy to see and approve what an agent did before it counts.

Read at TechCrunch · AI

Based on reporting from the original publisher. Visit the source for full context and later updates.

Publisher excerpt

More than 50% of managers say they're finding it harder to evaluate employees' work after adopting AI, according to a new survey.