--- title: "Managers Say They Use AI to Decide Who Gets… | Metaintro" canonical: "https://www.metaintro.com/blog/managers-using-ai-decide-who-gets-laid-off" language: "en" author: "drashtigarach" published: "2026-08-03T12:06:11.000Z" modified: "2026-08-03T12:06:11.000Z" --- [Back to Blog](/blog) [AI](/blog/tag/ai)[Layoffs](/blog/tag/layoffs)[Jobs](/blog/tag/jobs)[News](/blog/tag/news) # Managers Say They Use AI to Decide Who Gets Laid Off in 2026 59% of managers who use AI say they use it to decide who to lay off, and 38% were never trained on ethical AI in HR. What that means for your job security. [![Drashti Garach](https://cdn.metaintro.com/rs:fill:40:40/q:72/plain/images/5719d740-e510-42bc-8017-e040d145f35f_1766029465094.png)Drashti Garach @DrashtiGarach](/blog/author/drashtigarach) [August 3, 2026](/blog/archive/2026/08)13 min read ![Managers Say They Use AI to Decide Who Gets Laid Off in 2026](https://cdn.metaintro.com/rs:fill:1200:675/q:78/plain/images/managers-using-ai-decide-who-gets-laid-off-hero.png) [https://x.com/intent/tweet?text=Managers%20Say%20They%20Use%20AI%20to%20Decide%20Who%20Gets%20Laid%20Off%20in%202026&url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fmanagers-using-ai-decide-who-gets-laid-off](https://x.com/intent/tweet?text=Managers%20Say%20They%20Use%20AI%20to%20Decide%20Who%20Gets%20Laid%20Off%20in%202026&url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fmanagers-using-ai-decide-who-gets-laid-off)[http://www.facebook.com/sharer.php?u=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fmanagers-using-ai-decide-who-gets-laid-off](http://www.facebook.com/sharer.php?u=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fmanagers-using-ai-decide-who-gets-laid-off)[https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fmanagers-using-ai-decide-who-gets-laid-off&title=Managers%20Say%20They%20Use%20AI%20to%20Decide%20Who%20Gets%20Laid%20Off%20in%202026](https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fmanagers-using-ai-decide-who-gets-laid-off&title=Managers%20Say%20They%20Use%20AI%20to%20Decide%20Who%20Gets%20Laid%20Off%20in%202026)[mailto:?subject=Managers%20Say%20They%20Use%20AI%20to%20Decide%20Who%20Gets%20Laid%20Off%20in%202026&body=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fmanagers-using-ai-decide-who-gets-laid-off](mailto:?subject=Managers%20Say%20They%20Use%20AI%20to%20Decide%20Who%20Gets%20Laid%20Off%20in%202026&body=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fmanagers-using-ai-decide-who-gets-laid-off) Managers who use artificial intelligence at work are turning to it to help make staff cuts, with 59% saying they use AI when deciding who to lay off and 58% when deciding who to fire, according to a survey reported by [HR Dive](https://www.hrdive.com/news/managers-are-using-ai-to-make-layoff-decisions/826697/). The survey, conducted by ResumeTemplates.com, covered 1,000 US managers with direct reports, all of whom use AI at work. Note that qualifier, because it matters for how you read the numbers. At [Metaintro](https://www.metaintro.com) we cover this because most public discussion of AI and layoffs asks whether AI is causing job cuts. This asks something more immediate, which is whether AI is choosing who goes. ## What did the survey actually find? That the practice is widespread among AI-using managers rather than marginal. Fifty-nine percent said they use AI when deciding who to lay off and 58% when deciding who to fire. Beyond the headline, 1 in 4 managers said they use AI to help decide who gets laid off often or all the time, which describes a routine part of the process rather than an occasional experiment. The supervision picture is more reassuring, though not entirely. Fifty-seven percent of managers said they would never let AI make layoff decisions without supervising the process. But 43% said they occasionally take a hands-off approach, and 17% said they do so often or all the time. Ninety-one percent said they would override an AI layoff recommendation they disagreed with, which is the single most encouraging number in the study and also the one most dependent on a manager noticing there is something to disagree with. Confidence changes behaviour in a way that should give anyone pause. Among managers who said they were confident about AI's layoff decisions, 80% let it run unsupervised at least occasionally, compared with about half, 52%, of those who described themselves as only somewhat confident. Confidence, in other words, reduces oversight. That is the opposite of how a high-stakes decision process ought to behave, since confidence in a tool is not the same thing as the tool being correct. ## What inputs are managers feeding it? Mostly what you would expect, and one thing you would not. When managers do use AI to help decide layoffs, 80% ask it to weigh performance and productivity scores, 57% ask it to consider attendance, and 42% ask it to look at salary or cost. Those three are defensible inputs to a redundancy process, and they are broadly the criteria a human would apply. The outlier is the one worth reading twice. Almost one third, 31%, direct AI to consider frequent sick days or medical leave. That is nearly a third of managers asking a model to factor health-related absence into who loses their job. In many jurisdictions, selecting for redundancy on the basis of sickness absence or medical leave carries serious legal risk, particularly where that absence relates to a disability, a pregnancy or a protected condition. This is where the training gap becomes alarming rather than merely untidy. Thirty-eight percent of managers said they had never been trained on the ethical use of AI in making HR decisions. A manager who has not been trained is unlikely to know that a criterion feels reasonable and is legally hazardous, and a model asked to optimise against that criterion will do so efficiently and without objection. Our reporting on [how AI is being used in workplace decisions about pay and performance](https://www.metaintro.com/blog/amazon-walmart-workers-ai-hr-decisions-2026) covers the same pattern in adjacent processes. ## Should you assume an algorithm decided your layoff? No, and it is worth being careful here, because the finding is narrower than the headline suggests. The survey covered only managers who already use AI at work, so it does not tell you what share of all managers do this. It also measures self-reported use of AI as an aid, not autonomous decision-making. Ninety-one percent said they would override a recommendation they disagreed with, which describes a tool in the loop rather than a machine in charge. What the data does support is that AI is now a routine input into redundancy selection at a meaningful number of companies, and that the guardrails are uneven. The risk is not a rogue system firing people. It is a tired manager accepting a ranked list that looks objective, produced from criteria nobody examined, at a moment when they are under pressure to move quickly. Our analysis of [how one large employer described its own layoff selection process](https://www.metaintro.com/blog/2026-tech-layoffs-companies-cited-ai-tracker) shows how carefully companies word these things, and how little detail they usually volunteer. The practical consequence for an individual is limited but real. If you are laid off, you are unlikely to be told whether a model was involved. In several jurisdictions you can ask about the criteria used for selection, and that question is worth asking in writing regardless of what you suspect about the method. Criteria have to be defensible whether a human or a system applied them, and our guide to [applying for unemployment benefits promptly after a layoff](https://www.metaintro.com/blog/laid-off-2026-apply-unemployment-benefits-now) covers the steps that matter most in the first week. ## What should you do to protect yourself? Make your contribution legible in the data rather than only in people's memories. If performance and productivity scores are the input in 80% of cases, then whatever your organisation records as your performance is doing more work than your reputation is. That means closing out tickets properly, keeping your objectives updated, and ensuring completed projects are actually logged against your name. This is unglamorous administrative housekeeping and it is now closer to job security than it used to be, and it sits alongside the broader pattern we describe in [the AI job security blind spot facing tech workers](https://www.metaintro.com/blog/ai-job-security-blind-spot-tech-workers-2026). Pay particular attention to anything that could read as an absence pattern. Given that 57% of managers ask AI to weigh attendance and 31% ask it to consider frequent sick days or medical leave, the way your leave is categorised matters. Approved medical leave, parental leave and disability-related absence should be recorded as such rather than sitting in a generic absence field. If your records are wrong or vague, correcting them with HR is a small task with a disproportionate payoff, and it is far easier to do while employed. Understand your rights before you need them. Several jurisdictions restrict automated decision-making in employment, and workers in the European Union in particular have rights regarding decisions made by automated processing. If you are outside the US, your protections in this area are likely stronger than the survey's American framing suggests. Our reporting on [how workplace surveillance and monitoring are reshaping the way people work](https://www.metaintro.com/blog/ai-workplace-surveillance-breaking-how-we-work) covers the broader regulatory direction. Finally, keep this in proportion. The most common inputs are performance, attendance and cost, which are the same things managers weighed before any of this existed. AI has not introduced a new basis for selection so much as it has made the existing basis faster, more consistent and considerably harder to argue with. That last quality is the genuine change, because a number produced by a system carries an authority that the same number written on a whiteboard does not. ## Why does an AI-produced ranking carry more weight than it should? Because a number looks settled in a way a judgement does not. When a manager writes a list of names on a whiteboard, everyone in the room understands it is a set of opinions and treats it accordingly. When the same list emerges from a system, ranked and scored, it acquires an appearance of objectivity that the underlying reasoning has not earned. The inputs are still someone's choices about what matters. The output simply hides that. That is the mechanism to watch, and it is more important than any single statistic here. Forty-three percent of managers occasionally take a hands-off approach and 17% do so often or all the time, but even the supervising majority are reviewing something that arrives looking finished. Reviewing a proposal is psychologically different from constructing one, and people accept more when they are checking than when they are deciding. The 91% who say they would override a recommendation they disagreed with are telling the truth about their intentions, and the harder question is how often they notice grounds for disagreement in the first place. For an employee, this is why asking about criteria is more useful than asking about tools. Whether a model was involved is often unanswerable and sometimes irrelevant. What the selection criteria were, and whether they were applied consistently, is a question with a defensible answer that an employer is obliged to be able to give. Our reporting on [how workers should respond when a layoff announcement leaves out the detail that matters](https://www.metaintro.com/blog/how-leaders-stop-quiet-exodus-after-layoff-2026) covers how to frame that request. ## What should employers take from this? That the training gap is the fixable part and by far the most urgent. Thirty-eight percent of managers having no training in the ethical use of AI for HR decisions is not a technology problem. It is a governance failure with a straightforward remedy, of the kind we examine in [how silent layoffs spread through the tech industry](https://www.metaintro.com/blog/silent-layoffs-tech-industry-ai-impact), and it sits alongside the finding that a third of managers are feeding medical leave into layoff models. Those two numbers together describe meaningful legal exposure that most organisations have not priced. The confidence finding deserves specific attention from anyone designing these processes. If confident managers supervise less, then the people most at risk of accepting a flawed recommendation are precisely the ones least likely to question it. Any sensible control would therefore make review mandatory rather than discretionary, since relying on managers to opt into scrutiny selects against the cases where scrutiny is most needed. Our reporting on [why most laid-off workers end up accepting lower pay](https://www.metaintro.com/blog/57-percent-laid-off-workers-accept-lower-paying-job) shows what happens downstream when selection goes wrong. There is a reasonable defence of the practice worth acknowledging. Redundancy selection done by humans alone is not a model of fairness either, and it is subject to favouritism, recency bias and inconsistency between managers. A documented, consistent set of criteria applied uniformly can genuinely be fairer than a room of people making judgement calls. The question is not whether AI is worse than human decision-making in the abstract. It is whether this particular implementation, with these inputs and this level of oversight, is better than what it replaced. On a third of managers weighting sick leave, the answer in those cases is clearly no. Our reporting on [why manager training matters more than tooling in AI adoption](https://www.metaintro.com/blog/leadership-upskilling-beats-culture-change-2026) makes the same argument from the leadership side. ## Does this change how you should read a redundancy consultation? It should sharpen the questions you ask, without turning the process adversarial. Consultation is where selection criteria are supposed to be explained, and it is the one point at which you have a formal right to interrogate them. Asking what the criteria were, how they were weighted and over what period performance was assessed is entirely standard, and none of it requires you to raise the subject of AI at all. The period question is the most useful and the least asked. A model weighing performance and productivity scores is weighing whatever window it was given, and a strong employee coming off one difficult quarter can look very different across six months than across twenty-four. If the assessment window is short, that is worth challenging on its own terms, because it is a human choice about the input rather than anything the technology decided. Keep your own evidence in a form you can produce quickly. If selection weighted performance data, then your ability to show completed work, delivered outcomes and any recognition you received becomes directly relevant rather than merely reassuring. People in consultation frequently discover that everything supporting their case lives in systems they have just lost access to. Exporting a simple record while you still have a login is a small precaution that occasionally changes an outcome. ## People Also Asked ### Q: How many managers use AI to decide layoffs? A: Among managers who use AI at work, 59% said they use it when deciding who to lay off and 58% when deciding who to fire. One in four said they do so often or all the time. The survey covered 1,000 US managers with direct reports who all use AI at work. ### Q: What criteria do managers ask AI to consider? A: When using AI to help decide layoffs, 80% ask it to weigh performance and productivity scores, 57% ask it to consider attendance, 42% ask it to look at salary or cost, and 31% direct it to consider frequent sick days or medical leave. ### Q: Do managers check what the AI recommends? A: Mostly. 57% said they would never let AI make layoff decisions without supervising the process and 91% said they would override a recommendation they disagreed with. However, 43% occasionally take a hands-off approach, 17% do so often or all the time, and 38% have never been trained on the ethical use of AI in HR decisions. ## Related Articles - [How AI Is Reshaping HR Decisions on Pay](https://www.metaintro.com/blog/amazon-walmart-workers-ai-hr-decisions-2026) - [Tracker of Companies Citing AI in 2026 Layoffs](https://www.metaintro.com/blog/2026-tech-layoffs-companies-cited-ai-tracker) - [AI Workplace Surveillance Is Breaking How We Work](https://www.metaintro.com/blog/ai-workplace-surveillance-breaking-how-we-work) - [Leadership Upskilling Beats Culture Change](https://www.metaintro.com/blog/leadership-upskilling-beats-culture-change-2026) - [The AI Job Security Blind Spot for Tech Workers](https://www.metaintro.com/blog/ai-job-security-blind-spot-tech-workers-2026) - [The Real Reason Workers Push Back on AI](https://www.metaintro.com/blog/real-reason-workers-resist-ai-not-technology) - [Colorado AI Surveillance and Wage Setting](https://www.metaintro.com/blog/colorado-ai-surveillance-wage-setting) - [Emotion AI and White-Collar Workplace Surveillance](https://www.metaintro.com/blog/emotion-ai-white-collar-workplace-surveillance) - [Executives Expect AI Layoffs, Mercer Survey](https://www.metaintro.com/blog/99-percent-executives-expect-ai-layoffs-2028-mercer-survey) - [Laid Off in 2026, Apply for Unemployment Benefits Now](https://www.metaintro.com/blog/laid-off-2026-apply-unemployment-benefits-now) - [Most Laid-Off Workers Accept a Lower-Paying Job](https://www.metaintro.com/blog/57-percent-laid-off-workers-accept-lower-paying-job) - [The AI Layoff Myth and Companies Using AI as Cover](https://www.metaintro.com/blog/ai-layoff-myth-companies-using-ai-cover-job-cuts) Future-proof your career. When performance scores and attendance records are feeding directly into selection decisions, the most useful thing you can do is know what the wider market values in your skills rather than what one internal system records about them. [Metaintro](https://www.metaintro.com) matches your profile against roles worldwide so you can see your position independently of any employer's ranking, and you can [create your free account here](https://www.metaintro.com/signup) to start. ### Share this article [https://x.com/intent/tweet?text=Managers%20Say%20They%20Use%20AI%20to%20Decide%20Who%20Gets%20Laid%20Off%20in%202026&url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fmanagers-using-ai-decide-who-gets-laid-off](https://x.com/intent/tweet?text=Managers%20Say%20They%20Use%20AI%20to%20Decide%20Who%20Gets%20Laid%20Off%20in%202026&url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fmanagers-using-ai-decide-who-gets-laid-off)[http://www.facebook.com/sharer.php?u=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fmanagers-using-ai-decide-who-gets-laid-off](http://www.facebook.com/sharer.php?u=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fmanagers-using-ai-decide-who-gets-laid-off)[https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fmanagers-using-ai-decide-who-gets-laid-off&title=Managers%20Say%20They%20Use%20AI%20to%20Decide%20Who%20Gets%20Laid%20Off%20in%202026](https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fmanagers-using-ai-decide-who-gets-laid-off&title=Managers%20Say%20They%20Use%20AI%20to%20Decide%20Who%20Gets%20Laid%20Off%20in%202026)[mailto:?subject=Managers%20Say%20They%20Use%20AI%20to%20Decide%20Who%20Gets%20Laid%20Off%20in%202026&body=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fmanagers-using-ai-decide-who-gets-laid-off](mailto:?subject=Managers%20Say%20They%20Use%20AI%20to%20Decide%20Who%20Gets%20Laid%20Off%20in%202026&body=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fmanagers-using-ai-decide-who-gets-laid-off) For job seekers ## The job you want exists. We’ll find it. No more scrolling job boards. Get matched roles by text, with the pay and the reason they fit. [Get Started Free](/signup)[Search matching jobs](/jobs/search) - MatchCompare live roles against your current evidence. - PositionTurn proof projects into role-specific applications. - ImproveUse market feedback to keep the skill plan current. [Return to navigation](#main-navigation)