AI is changing employee reviews and adding work for managers

The review problem

94% of survey respondents want to improve their work with AI, and many believe they know where it can help most. Michael Levy, Monday.com’s head of market research and customer insights, told TechCrunch that the enthusiasm resembles the “magic” people saw in the early years of the internet in the 2000s.

But 78% said they had changed how they evaluate employees after adopting AI. Managers now pay more attention to whether people can write prompts, choose tasks to hand off and try new ways of working.

That creates a difficult question: how can managers assess an employee’s contribution when they cannot easily tell how much of the work AI did? The problem grows when managers use AI to make the assessments themselves. In the survey, 46% said they use AI to make decisions, while 39% said they ask it to draft employee reviews.

Research cited in the report points to risks. A study from the Luxembourg Institute of Socio-Economic Research, published earlier this year, found that when objective standards are unclear and people are assessed independently, large language models tend to produce ratings similar to human ones. A research paper from early 2025 also reported that AI models can show substantial bias against Black male candidates, with the direction and strength of bias varying across social groups.

Asked whether AI involvement in employee evaluations could introduce outside bias, Levy said companies had not yet worked out how to address the issue.

More tools, more supervision

AI is also changing the day-to-day work of managers. They must decide which tasks to delegate to agents, check the results and step in when something goes wrong. The report describes that oversight as a potential new layer of “invisible work.”

The software environment is getting harder to manage, too:

Levy said companies expected AI to help reduce the number of software tools they use. Instead, they have added applications and integrations, while employees have started using unofficial AI tools outside company-approved systems. Managers had expected to consolidate tools at the end of 2025, but more than 50% said they use them more often than a year earlier, and that growth is continuing.

The agent accountability gap

The oversight challenge is becoming more pressing as companies adopt agents that can act autonomously. In the survey, 58% of managers expected to supervise more AI agents than human employees over the next year, and 46% feared losing control over what those agents do.

Another 35% said they worried about being held responsible for decisions made by AI. Levy said managers did not want agents to be fully autonomous: for important tasks and key decisions, they want AI to involve and consult people.

I think the tension is not simply that AI makes management harder. It is that companies are changing how work is done and evaluated faster than they are building the systems to govern it. Until those catch up, the productivity gains may arrive with a growing bill for human supervision.