The intersection of artificial intelligence and human resources has been a story of promise—faster hiring, bias-free screening, data-driven performance reviews. But a recent survey reveals a far more unsettling application: a significant number of managers are now turning to AI models to make layoff decisions. The practice, which few employees are aware of, raises profound questions about fairness, transparency, and the future of work itself. As economic uncertainty prompts more companies to reduce headcount, the quiet introduction of algorithmic terminations threatens to erode trust and expose businesses to new legal perils.

A Glimpse Into the Algorithmic Axe

A newly published survey of managers across various industries found that a notable portion have used AI tools to identify which employees should be let go during restructuring efforts. The parameters fed into these systems are often deeply personal: attendance records, including sick days; chronological age; length of tenure; and even performance ratings that may themselves be influenced by biased human judgment. Some models go further, analyzing internal communication patterns or project management activity to estimate an individual’s “impact.” The results are then presented, sometimes as a simple list, guiding a manager’s hand in a decision that will upend a person’s livelihood.

The shift is happening without much fanfare. Many companies have quietly integrated AI into their HR suites, and the leap from using algorithms for recruitment to wielding them for dismissal is technically short. But ethically, it is a chasm. While proponents argue that AI can remove emotional bias and ground decisions in objective data, critics counter that the data itself is often a mirror of existing inequities—and that offloading accountability to a machine is a dangerous abdication of human responsibility.

How the Algorithms Are Reaching Their Verdicts

Understanding what happens inside these “black box” systems is crucial. Most workforce AI tools are trained on historical data. If a company’s past layoffs disproportionately affected older workers or those with intermittent health issues, the model may learn to replicate that pattern, encoding it as a feature of “low value.” Sick days, for instance, can be interpreted as a proxy for unreliability, ignoring the fact that chronic illness or family care obligations fall heavily and unevenly across the workforce. Similarly, tenure-based cuts can appear neutral but may systematically remove employees who have accumulated institutional knowledge and whose salaries reflect years of service rather than performance.

Then there is the morale impact on survivors. While some believe a data-driven process feels more objective, the opacity of AI decisions can breed suspicion. An employee terminated without a clear, human-delivered rationale may suspect discrimination, leaving the company exposed to wrongful termination lawsuits. In jurisdictions with strong worker protections, such as several European countries, the use of automated individual decision-making without meaningful human intervention is already tightly regulated or outright restricted. The EU’s AI Act, for example, classifies employment-related AI systems as high-risk, requiring conformity assessments and transparency obligations that many ad-hoc AI layoff tools likely fail to meet.

The Ethical and Legal Minefield

Experts warn that this trend is a ticking time bomb for HR departments. The lure of efficiency is understandable—a model can analyze thousands of employee profiles in minutes, offering a seemingly clean solution to a painful task. Yet, the absence of human oversight in such high-stakes decisions violates basic principles of procedural fairness. Employees have a right to understand the criteria used to end their employment and to challenge flawed data. When the decision-maker is a neural network, neither is possible.

Legally, the risks are stark. Anti-discrimination laws in the United States, such as the Age Discrimination in Employment Act and the Americans with Disabilities Act, do not pause simply because a machine made the call. If an AI model’s reliance on sick leave or age disproportionately impacts a protected class, the employer is liable—even if the vendor provided the tool. Class-action lawsuits and regulatory investigations are a near certainty as awareness grows. Already, the Equal Employment Opportunity Commission has signaled that algorithmic bias in employment will be a priority area of enforcement.

Beyond litigation, brand damage can be severe. In a connected world where employees share experiences on platforms like Glassdoor and LinkedIn, a reputation for impersonal, AI-driven firing can deter top talent. The very engineers and knowledge workers companies hope to attract are often the most vocal critics of unaccountable technology.

Restoring Human Judgment at the Breaking Point

The solution is not to abandon AI in HR altogether—it has genuine value in surfacing trends, flagging anomalies, and reducing administrative grunt work. Rather, the case demands a strict boundary: algorithms should inform, never decide, when it comes to employment endings. Every layoff decision must involve a trained human manager who can interpret the data, question its sources, and apply context that no machine can grasp, such as an employee’s recent personal struggles, a temporary dip in performance due to project shifts, or the unique value of a team player who quietly holds the culture together.

Organizations should also conduct rigorous audits of any AI tool that touches workforce decisions, testing for disparate impact and ensuring that the model’s training data is vetted. Transparency reports—shared with employees or works councils—can help build trust, showing that technology is an assistant, not an executioner. In a hybrid and remote work era, where face-to-face conversations are rarer, keeping the human element front and center in difficult moments is even more critical. Some modern workforce platforms are starting to emphasize fairness and contextual data over simplistic metrics, providing managers with a fuller picture rather than a single score.

The challenge is ultimately cultural. No dashboard or predictive model can replace the empathy and accountability that a leader must bring to the hardest part of the job. As AI capabilities expand, the temptation to delegate painful tasks will grow. But if companies want to sustain engagement and trust through cycles of change, they must keep the ultimate responsibility in human hands—where it has always belonged.

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