AI performance evaluation has become harder for more than half of the 1,701 directors and senior managers in a new Nielsen survey commissioned by Monday.com, because finished work no longer shows how much of it the employee did. The same study found that 39% of those leaders now use AI to draft performance reviews, which puts AI on both sides of the assessment.
AI Performance Evaluation Got Harder Because Output No Longer Shows Effort
Monday.com published the study as “AI at work 2026: Squinting in the Sun,” covering senior leaders across seven countries. TechCrunch’s Ram Iyer reported its central finding on evaluation: work now reaches a manager’s inbox polished, and the manager cannot easily tell which parts came from the person and which from a model. Nearly half of the managers said they now spend more time reviewing and guiding their teams’ output than before. The review step has grown, because the finished artifact has stopped being reliable evidence of the skill that produced it.
78% of Managers Have Changed What They Grade Employees On
Managers are adjusting. According to TechCrunch’s account of the survey, 78% said they have changed how they evaluate employees since adopting AI. The new weight falls on how well staff write prompts, decide which tasks to hand to AI, and try new ways of working. Monday.com’s own summary adds that 85% of respondents believe being a good manager now requires different skills. Only 65% said their company has given them enough AI training, so many of the people redesigning evaluation criteria are doing it without formal preparation.
39% of These Leaders Now Use AI to Draft Performance Reviews
The survey found that 46% of managers use AI to help inform their decisions and 39% use it to draft performance reviews. That closes a loop. An employee’s work may be partly AI-generated, the manager cannot fully separate the two, and the judgment of that work may also be partly AI-generated. TechCrunch pointed to a 2025 study in PNAS Nexus that found large language models showed significant biases across social groups when evaluating people. Michal Miller Levi, Monday.com’s head of market research and customer insights, told TechCrunch that companies have not yet worked out how to handle that risk. Her broader diagnosis was that “AI is just moving faster than the organizational processes.”
Managers Report Longer Hours and More Tools Since Adopting AI
The extra review time sits inside a wider pattern in the data. Monday.com reports that 29% of respondents say AI has increased their working hours, and 59% say finding and testing where AI adds value takes more effort than it saves. Mandates have moved faster than infrastructure: 72% work at companies that require or strongly encourage AI use, while 29% describe their AI setup as fully optimized and centralized. TechCrunch reported that 57% now have more tools in circulation than before the AI rollout.
Managers Expect to Supervise AI Agents Within a Year
The evaluation problem is about to widen. According to TechCrunch, 58% of the managers expect to oversee more AI agents than human employees within the next year. 46% worry about losing control over what those agents do, and 35% fear being held responsible for decisions an AI made. Levi said managers want agents to consult people on crucial tasks and decisions instead of acting fully on their own.
Why This Matters for Engineering Leadership
Engineering managers face this problem in its sharpest form, because their teams’ output is code, and a pull request does not record how much of it a model wrote. The survey shows managers quietly rewriting their criteria and partly outsourcing the verdict to the same class of tools. Both moves can happen without anyone deciding them. Engineering leaders should make the decision explicit. Write down what the organization now evaluates, whether that is design reasoning, review quality, incident judgment or the choice of what to delegate to a model, and tell engineers before the review cycle starts. Set a rule on AI-drafted reviews: a model can help with wording, but every claim about a person needs evidence the manager gathered and can show. Budget the extra review time the survey describes instead of assuming AI has freed it. If 58% of these managers are right about supervising agents next year, the criteria written now will also become the template for judging agent work.
FAQ
Does AI make employee performance harder to evaluate?
More than half of the 1,701 directors and senior managers in a 2026 Nielsen survey for Monday.com said it does. Polished AI-assisted output makes it hard to tell how much of the work the employee did.
How many managers use AI to write performance reviews?
39% of the leaders surveyed said they use AI to draft performance reviews, and 46% use AI to help inform their decisions, according to TechCrunch’s reporting on the study.
How are managers changing how they evaluate employees after AI?
78% said they have changed their evaluation approach, giving more weight to how well staff write prompts, choose tasks to hand to AI and try new ways of working.
Who conducted the Monday.com AI at work 2026 survey?
Nielsen conducted it for Monday.com, polling 1,701 directors and senior managers across seven countries. Monday.com published it as “AI at work 2026: Squinting in the Sun.”
Is AI saving managers time?
Not consistently. 29% said AI has increased their working hours, nearly half spend more time reviewing their teams’ output, and 59% said finding where AI adds value takes more effort than it saves.