If your employer is running a headcount review right now, there is a real chance an AI tool is helping shape the outcome, and you probably have no idea it is happening. A ResumeTemplates.com survey of 1,000 US managers who use AI at work found that 59% now rely on AI when deciding who to lay off, and 58% do the same when deciding who to fire. Roughly 1 in 4 said they use it that way often or all the time. The deeper problem is not the tool itself. It is what the tool is being asked to weigh, and who is watching when it runs.

43% of managers said they occasionally take a hands-off approach to AI layoff decisions, and 17% said they do so often or all the time. More striking still: 38% of managers said they had never received training on the ethical use of AI in HR decisions. So the software is active, the oversight is thin, and the worker whose job is at stake typically has no visibility into any of it. For anyone navigating a performance review cycle or a restructuring announcement this year, that combination is worth understanding clearly.

What the AI is actually scoring you on

When a manager feeds a layoff question into an AI tool, they are typically asking it to rank employees on a bundle of workplace signals. According to the ResumeTemplates.com survey, 80% of managers direct AI to weigh performance and productivity scores, 57% factor in attendance, and 42% include salary or compensation cost. Tenure registers at 32% and paid time off at 23%.

Those inputs are broadly what most workers would anticipate. The next tier is where US employment law becomes directly relevant. 31% of managers direct AI to consider frequent sick days or medical leave in layoff decisions, and 14% include age. Julia Toothacre, chief career strategist at ResumeTemplates.com, put the legal stakes plainly: sick days, medical leave, and age stand apart from the factors a layoff usually turns on, because discrimination based on age, disability, or protected medical leave is illegal.

In the US context, this is not a hypothetical concern. The Americans with Disabilities Act (ADA) prohibits employers from taking adverse employment actions against workers because of a disability or a record of disability. The Age Discrimination in Employment Act (ADEA) protects workers aged 40 and over from employment decisions driven by age. The Family and Medical Leave Act (FMLA) protects eligible employees who take leave for qualifying medical or family reasons from retaliation or termination tied to that leave. The moment any of those characteristics become inputs in a scoring model, they can shape the output regardless of whether the manager intended them to. That is the mechanism the law targets, not just the intent.

Nobody can verify what the tool weighs

The most unsettling finding in the survey is not what the AI is doing. It is what no one can confirm. 58% of managers said they cannot verify whether their company has tested the AI they use for bias. Another 23% said it was not tested at all. Put together, more than 8 in 10 managers using AI for layoff calls are working with a tool whose fairness has either not been audited or cannot be audited.

Julia Toothacre of ResumeTemplates.com said the combination of untrained managers and untested tools creates a compounding problem: when the company cannot confirm the tool was tested for bias, there is no way to know what it weighs or whether the decision is defensible, and that gap is precisely where discrimination and wrongful-termination claims begin.

This gap is not just a compliance issue for employers. For you, it has a practical meaning in a termination meeting. If you are laid off and you believe a protected characteristic influenced the outcome, the employer may be unable to produce a clear, auditable rationale. The Equal Employment Opportunity Commission (EEOC) has made clear that algorithmic tools used in employment decisions are subject to existing anti-discrimination statutes, including Title VII of the Civil Rights Act of 1964, the ADA, and the ADEA. The agency does not offer a separate exemption because a machine generated the score. It is also worth knowing that some states and cities are moving beyond federal law: New York City's Local Law 144 requires employers using automated employment decision tools to conduct annual bias audits and notify candidates, and Colorado's AI Act, signed in 2024 and taking effect in 2026, imposes algorithmic impact assessment requirements on high-risk AI systems, including those affecting employment decisions. The federal floor is significant; the state and local ceiling can be higher depending on where you work.

The question of whether AI can do your job

There is a second, quieter use of AI in these decisions. Among managers who use AI for layoffs, 34% have asked the tool to assess whether a given person's role could be performed by AI instead, which amounts to about 1 in 5 of all managers surveyed. Separately, 44% of managers said they were asked to make that call themselves, and 75% of those concluded the role could be replaced.

Treat that 75% figure with caution before internalizing it. Research from PYX Labs, a research unit sponsored by Perceptyx, found that AI struggled with nuance and complex results, particularly when interpreting open-ended employee feedback. At the Society for Human Resource Management's SHRM26 conference in June, a leadership expert observed that AI tends to focus on what already exists in a record rather than a worker's potential. The practical implication: a model is reasonably good at cataloguing the measurable outputs a role produced last quarter. It is poor at capturing the judgment calls, stakeholder relationships, and institutional knowledge that made those outputs possible and correct.

If a role near yours has been eliminated on the basis that AI can handle the work, the honest follow-up question is whether AI can handle the observable, recorded outputs. It almost certainly cannot replicate the unrecorded work that kept those outputs from being wrong.

What to do if you suspect an AI-influenced decision

There is no reliable way to confirm from the outside whether a specific layoff or termination was AI-assisted. What you can do is pay close attention to the signals around you and raise the right questions early, before a decision is finalized.

• Watch what is being tracked. If a performance improvement plan, restructuring memo, or review cites metrics you were not aware of, such as login frequency, message volume, keystroke data, or a productivity score, that is a signal an automated system is involved somewhere upstream in the process.
• Ask directly in the meeting. In a termination or restructuring conversation, you are entitled to ask whether any AI or automated tool informed the decision and, if so, what factors it considered. The response, or the refusal to respond, is informative either way.
• Document protected status in writing. If you recently took FMLA leave, disclosed a disability covered by the ADA, filed a workers' compensation claim, or are 40 or older, note these facts in writing during the process. An employer's obligation to justify a decision on non-protected grounds is considerably stronger when the record shows they had prior knowledge of those characteristics.
• Request the selection criteria. In structured layoff processes, you can ask for the criteria used and how you were scored against them. If attendance or sick leave appear as factors and you took protected leave, that discrepancy is worth raising with an employment attorney or filing as a charge with the EEOC, which processes complaints at no cost to the worker.
• Build your own evidence file. Copies of positive feedback, completed deliverables, client communications, and any metrics your manager has commended in writing are your counter-record. If the AI's picture of your contributions is thinner than the reality, your own documentation fills in what the model missed.

The one figure that points toward accountability

The survey did surface one number that cuts against full pessimism. 91% of managers said they would override an AI layoff recommendation they disagreed with, and 57% said they would never allow AI to make layoff decisions without supervising the process. Full automation is not the stated preference of most managers in this data set.

The complication is how confidence interacts with oversight. Among managers who described themselves as confident in the AI's layoff decisions, 80% allowed it to run unsupervised at least occasionally, compared with 52% of those who called themselves only somewhat confident. The more a manager trusts the tool, the less they interrogate it. That is the inverse of how meaningful supervision is supposed to function, and it is the gap that creates exposure for workers and legal liability for employers.

What this means for your next role or your current one

If you are employed right now, the useful response is not to optimize your behavior for an algorithm you cannot see. It is to make sure the record that exists around you, in your manager's inbox, in project documentation, in cross-team communications, reflects the full scope of what you actually do. Regular one-on-one conversations where you name specific deliverables and outcomes are more protective than a single polished annual review, because any AI reading your file is reading the ongoing signal, not just the formal snapshot.

If you are exploring a new role, the ResumeTemplates.com data is a reasonable prompt to add a few questions to your final-round interviews. Asking whether the company uses AI in performance management or workforce planning, whether those tools have been audited for bias, and who reviews the output before a decision is made will tell you something real about how the organization thinks about accountability. Employers who have worked through these questions will have a clear answer. Employers who have not will signal that too, through the way they respond.

The EEOC, the DOL (Department of Labor), and a growing number of state agencies are actively sharpening their guidance on algorithmic employment decisions. The legal landscape is moving. Until the gap between tool deployment and manager training closes, the most useful thing to understand is that the person making the call about your job may be looking at a score, and in too many companies right now, nobody, including them, can fully explain how it was produced.