The use of artificial intelligence in human resources has moved far beyond résumé screening and chatbots. A recent survey has revealed a sobering new frontier: managers are increasingly turning to AI models to determine who to let go during workforce reductions. This shift, while perhaps intended to bring objectivity to one of the hardest decisions a leader can make, raises profound questions about fairness, legality, and the erosion of managerial judgment.

How the Survey Exposed a Growing Trend

The data, drawn from a poll of managers across multiple industries, indicates that a significant minority are already experimenting with AI tools to guide layoff decisions. The systems are not just spitting out performance metrics; in some cases, they are being asked to weigh factors such as sick-day usage, employee age, and tenure. This means the algorithm could effectively penalize a worker for having taken medical leave or for being closer to retirement, without understanding the human context behind those data points.

Proponents of the approach might argue that it removes personal bias and makes the process more data-driven. If a company must cut 10% of its workforce, wouldn’t it be fairer to let a machine identify which roles are least critical based purely on numbers? That logic, however, ignores the fact that AI models are trained on historical data that may already embed systemic inequalities. When sick leave or age become decision variables, the tool can inadvertently amplify discrimination rather than eliminate it.

What’s Actually Being Fed into the Algorithm

The survey’s most troubling finding is the type of data points managers are willing to feed into these layoff models. Beyond standard productivity metrics, the list includes attendance records, years of service, and even inferred characteristics like potential future attrition risk. An employee who took multiple short-term disability leaves for a chronic illness might be flagged as a “high-risk” retention case—not because of their performance, but because of a pattern the AI interprets as costly.

This creates a dangerous feedback loop. If a model learns that workers who take more sick days eventually leave the company, it may advocate for preemptively removing such individuals, thereby turning a correlation into a self-fulfilling prophecy. Moreover, managers themselves may not fully understand how the AI weighs these inputs. Many AI systems are black boxes, offering little transparency into why one employee was selected over another. For the person whose livelihood is terminated, that lack of explainability can feel like a violation of basic dignity.

The Legal and Ethical Tightrope

Employment law in many jurisdictions is clear: layoff decisions must not be discriminatory. When age or health-related absences factor into an algorithmic decision, employers may inadvertently violate laws protecting workers over 40, employees with disabilities, or those on protected leave. Even if the AI does not explicitly use a protected category, the proxy variables it relies on—such as tenure as a stand-in for age—can produce the same discriminatory effect.

Regulators are starting to pay attention. The EU’s AI Act, for example, already classifies AI systems used in employment decisions as high-risk, requiring rigorous documentation, human oversight, and transparency. Companies that offload layoff decisions to opaque algorithms without these safeguards could soon find themselves facing investigations, fines, and reputational damage. Even in regions with less prescriptive legislation, the risk of class-action lawsuits looms large. A single disgruntled former employee who can demonstrate that an algorithm used age-correlated data could trigger a costly legal battle.

Beyond the courtroom, there is the matter of trust. When news spreads that a company used an AI to choose who to fire, surviving employees may feel they are working under a disembodied surveillance system rather than human leaders. Engagement, already fragile after any layoff round, can plummet. The short-term operational gain of a “data-driven” cut may be erased by the long-term cultural damage.

Where Human Judgment Must Prevail

None of this is to say that data has no role in workforce planning. Effective HR analytics can help identify structural overstaffing, flag departments with disproportionate costs, or model the impact of different scenarios on business continuity. That is a far cry, however, from letting a machine make individual termination decisions. The nuanced reality of an employee’s contributions—their mentorship of junior colleagues, their institutional knowledge, their ability to steady a team during a crisis—rarely shows up in a spreadsheet.

Seasoned HR professionals know that a thoughtful layoff process requires more than a ranking algorithm. It demands calibration by managers who can weigh potential, passion, and cultural fit alongside hard metrics. It requires compassion, clear communication, and a sense of fairness that an AI simply cannot replicate. Technology can assist in anonymizing initial data sets to identify broad patterns, but the final, painful choices must rest with people who can be held accountable.

As companies navigate an increasingly uncertain economic landscape, the pressure to act quickly and appear data-savvy is immense. Yet the appeal of an automated, “bias-free” layoff solution is a mirage. It substitutes a veneer of objectivity for genuine fairness and exposes the organization to a host of new risks. For those committed to doing right by their teams, the antidote is not to ban AI from the HR department altogether, but to draw a hard line: algorithms may inform, but they must never decide who stays and who goes.

Originally published by XMF, inspired by publicly reported industry news.