Why Companies Lay Off Employees Under the AI Pretext—And Why Some Get Called Back

The irony is stark: companies aggressively lay off employees claiming AI will replace them, only to quietly rehire those same workers weeks later. This pattern reveals something uncomfortable about how corporate leadership uses artificial intelligence as cover for what is fundamentally a cost-cutting decision.

The Immediate Contradiction: Lay Off Today, Rehire Tomorrow

The headlines told one story. In late February 2025, Jack Dorsey announced that his fintech company Block had laid off more than 4,000 employees in a single sweep, slashing its workforce from 10,000 to below 6,000. The justification was straightforward: “AI tools have changed everything.” The message was clear—these roles would be absorbed by artificial intelligence.

But the follow-up told a different story. Within weeks, laid-off employees started receiving calls back to the office. According to reports from major business outlets, these recalls came from multiple departments: engineering, recruitment, and beyond. Some were told they’d been wrongfully terminated due to “clerical errors.” Others revealed that managers had continuously advocated for their rehiring. Several received unexpected calls, no explanation offered, simply asking them to return to work.

This pattern isn’t new. In 2022, when Elon Musk acquired Twitter and immediately laid off roughly half the workforce (over 3,000 people), he quietly rehired dozens of them after realizing that certain key positions simply cannot operate without human decision-making. More recently, Klarna—the Swedish payment company that publicly celebrated laying off over 1,000 employees by claiming AI customer service could replace 700 human agents—admitted by mid-2025 that they had “moved too fast” and began selectively rehiring customer service staff.

The question becomes unavoidable: if AI was truly ready to replace these workers, why do companies need to bring them back?

The Economics of Replacement: Why AI Isn’t Cheap

The answer lies in something that rarely makes headlines: the actual operational cost of AI deployment.

Enterprise-level AI doesn’t run on magical efficiency. It runs on tokens, and tokens are expensive. Claude 3.5 Opus charges $5 per million input tokens and $25 per million output tokens. For comparison, domestic large language models offer lower rates—Alibaba’s Qwen 3.5 Plus costs approximately 0.8 yuan ($0.11 USD) per million input tokens and 4.8 yuan ($0.66 USD) per million output tokens—but the overhead remains substantial.

Consider a real example: an experienced user working with Claude 4.5/4.6 for routine investment research and lifestyle queries burned through approximately $6,000 worth of tokens in just over one month. That’s $72,000 annually for a single individual using AI as an assistant.

Now scale that to enterprise level. A good-looking college graduate in regions with educational inflation might be hired as a customer service representative for roughly 3,000 yuan ($414 USD) per month. But training an AI customer service system to genuinely handle complex tickets, access multiple knowledge bases, conduct multi-turn conversations, and maintain stable uptime? That investment dwarfs a single employee’s salary. The computational infrastructure, continuous refinement, knowledge base integration, and error handling all add up to costs that make the $3,000/month human employee look like a bargain.

This is why companies lay off employees and then bring some back. The initial decision to lay off was often made by executives focused on headcount reduction as a quick fix for budget pressures. The rapid rehiring reveals the truth: certain roles genuinely require human judgment, accountability, and adaptability that no current AI system can replicate at a price point that justifies replacement.

When Efficiency Improvements Become Hidden Burdens

Even when AI doesn’t replace workers entirely, it creates a different problem rooted in what economists call the Jevons Paradox. The concept is simple: efficiency improvements don’t lead to reduced consumption of a resource; instead, they lead to increased total usage because improved efficiency lowers costs and expands demand.

In the workplace during the AI era, this plays out as follows: as AI tools improve employee output capacity, management doesn’t grant workers more leisure time. Instead, companies demand that employees complete significantly more work in the same time period.

So-called “productivity gains” become disguised workload increases. The narrative that AI liberates human labor is, at its core, a mischaracterization. What actually happens is that remaining employees take on expanded responsibilities. They learn to use AI tools, integrate them into workflows, troubleshoot when systems fail, and ultimately produce more output while receiving no corresponding increase in compensation or relief from existing duties.

The Organizational Damage That AI Cannot Compensate For

There’s a dimension to corporate layoffs that purely technical thinking misses: organizations are fundamentally human entities. Where humans organize, informal networks emerge—relationships that drive collaboration, knowledge-sharing, mutual support, and institutional wisdom.

Companies can integrate AI into formal organizational structures. They cannot integrate AI into the informal structures that actually make workplaces function. When layoffs occur—whether justified or not—companies don’t just cut labor; they cut organizational muscle. They eliminate people who served as unofficial mentors, problem-solvers, relationship-builders, and institutional memory keepers.

The remaining workforce bears not only heavier workloads but also the psychological weight of uncertainty, diminished collaboration, and increased individual accountability. There are fewer colleagues to delegate to, fewer people to absorb blame, and fewer channels through which problems can be solved informally.

Why Smart Leaders Choose Expansion Over Layoffs

During NVIDIA’s GTC 2026 conference, CEO Jensen Huang made a pointed critique of companies laying off staff under the guise of AI advancement. His words were direct: “Those leaders who rely on layoffs to cope with AI are doing so because they can’t think of better solutions. They’ve exhausted their creative ideas. Even with the strongest tools available, they won’t use them to expand.”

Jensen’s observation cuts to the heart of it. AI is a multiplier—it can expand productive capacity, enable new business lines, and create opportunities that didn’t previously exist. Leaders who understand this use AI to hire more strategically, not to eliminate headcount. They recognize that the real competitive advantage lies in combining powerful tools with expanded human capability, not in replacing human capability wholesale.

The companies that are truly thriving in the AI era aren’t the ones that lay off the most employees. They’re the ones that layer AI into their existing workforce and then hire more aggressively to pursue new directions.

The Pattern Reveals the Real Motivation

When you examine the sequence—aggressive layoffs announced with great fanfare, followed by quiet rehiring within weeks—a clearer picture emerges. The public layoffs serve a purpose: they signal to the market that management is “adapting to AI,” demonstrating decisive action and cost-consciousness. Investors appreciate visible headcount reduction; it looks like efficiency.

The rehiring, by contrast, happens quietly, with minimal press coverage. Managers advocate internally for their best people to be rehired. HR ushers returning employees back through the door with vague explanations. The company avoids the headlines that would undermine the narrative established by the layoff announcement.

In reality, the decision to lay off employees has little to do with whether AI can actually replace those roles. It has everything to do with cost reduction in the short term, executive pressure to show quick action, and the political reality that announcing layoffs plays better in business media than announcing measured, strategic growth.

What This Means for Workers and the Future of Work

The fact that companies lay off employees and then rehire them within weeks doesn’t indicate that leadership has come to its senses or that the AI revolution is a false alarm. Instead, it reveals that we’re in an awkward transition period where executives are making rushed decisions based on incomplete information and market pressure.

AI will indeed change many professions and skill sets. What the rehiring pattern demonstrates is that this change is neither magical nor immediate. Companies that laid off staff hoping AI would seamlessly absorb the workload discovered they were wrong. But rather than admit this clearly, they quietly bring people back and continue pretending the strategy was sound.

Meanwhile, the employees caught in this cycle absorb the real damage. They’ve experienced the stress of job loss, the uncertainty of potential permanent displacement, the public humiliation of being declared redundant, and then the administrative awkwardness of being rehired under vague circumstances. The psychological toll cannot be erased by a return phone call.

The future of AI in the workplace will likely be defined not by sudden job elimination but by continuous pressure on remaining workers to do more, earn the same, and adapt endlessly to new tools and methodologies. Companies won’t openly admit this because it’s a harder sell than “AI changes everything.” But the quiet rehiring of recently laid-off employees suggests that many executives are beginning to understand it, even if they won’t say it out loud.