The Great AI Layoff Paradox: Why Companies Keep Rehiring Despite AI Claims

The corporate world has entered a peculiar phase where artificial intelligence serves as both the justification for mass dismissals and the unstated admission of their failure. When Block announced its intention to lay off over 4,000 employees in February 2025—slashing its workforce from 10,000 to under 6,000—founder Jack Dorsey declared it a necessity because “AI tools have changed everything.” Yet within weeks, those same employees began receiving calls asking them to return. This wasn’t a mistake in the headline; it was a crack in the logic.

The Quick U-Turn: How AI Layoffs Became Employee Recalls

The rehiring wasn’t subtle. According to Business Insider, employees from engineering, recruitment, and design departments were brought back through various channels. Some received messages explaining their dismissal was due to a “clerical error.” Others had managers who personally advocated for their return. A few got calls a week after their pink slips, with no explanation offered—just an invitation to come back. The pattern was clear: the company had calculated wrong.

What Block’s experience reveals is that the decision to lay off staff based on AI capability wasn’t grounded in reality. Certain positions and workflows simply cannot be automated away with a command. Knowledge transfer, institutional memory, and specific expertise don’t evaporate because a new technology exists. The return of these employees suggests something uncomfortable: Block needed them back because AI didn’t fill the gap as promised.

The Vanishing Economics: Why AI Isn’t the Cheap Labor Replacement

The fundamental issue plaguing companies isn’t that AI can’t do the work—it’s that AI remains extraordinarily expensive to operate at scale. A single month of intensive AI usage can cost more than a year of mid-level human labor. Claude Opus 4.6 charges $5 per million input tokens and $25 per million output tokens. Domestic alternatives are cheaper but still significant: Qwen 3.5 Plus costs 0.8 yuan per million input tokens and 4.8 yuan per million output tokens. For context, one user’s month-long experiment with OpenClaw consumed approximately $6,000 in token costs. That sum could hire a capable professional for several months in most regions outside major Western cities.

The math becomes even more unfavorable when considering integrated enterprise AI systems. Replacing a customer service department doesn’t mean deploying ChatGPT and calling it done. It requires building systems that handle complex tickets, access multiple knowledge bases, maintain context across conversations, and operate without downtime. That infrastructure costs far more than the $3,000 monthly salary of an educated customer service representative.

Klarna discovered this reality in 2024 when it announced AI would handle the workload of 700 customer service agents after the company laid off more than 1,000 people. By May 2025, Klarna began rehiring customer service staff, with leadership admitting they had “moved too fast.” The company had learned what Block was learning: the gap between AI capability in marketing materials and AI capability in production systems remains vast.

The Burden Disguised as Liberation: Jevons Paradox in the Workplace

There’s an economic principle called the Jevons Paradox that explains why efficiency improvements don’t always reduce resource consumption. Instead, lower costs and increased capability often drive higher total usage. Apply this to the workplace, and the picture becomes darker.

When companies integrate AI and employee productivity theoretically increases, companies don’t grant employees more free time. They demand more output. The employees who remain after a layoff don’t experience liberation—they experience intensification. They work with AI as a tool not to work less, but to do more in the same hours. This is efficiency rebranded as exploitation.

The workplace version of Jevons Paradox reveals that AI doesn’t reduce work burdens; it transforms them. Tasks expand to fill the capacity created by new tools. Employees aren’t freed; they’re simply expected to produce more, faster, with the same compensation. The promise of AI liberating human labor has revealed itself as marketing fiction.

Why Organizational Structure Can’t Be Automated Away

Companies are not just collections of tasks. At their core, they are human systems with formal structures and invisible networks. These invisible networks—the informal relationships, the trust hierarchies, the unwritten protocols—are impossible for AI to replicate or replace.

When a company lays off staff because “AI changes everything,” it doesn’t just cut headcount. It damages organizational tissue. The remaining employees lose collaborators, lose people who served as buffers, lose colleagues who absorbed certain interpersonal complexities. More critically, they lose scapegoats—people who could share responsibility, who could shoulder blame, who could distribute organizational friction.

The employees who stay behind don’t just work harder; they absorb the anxiety, risk, and responsibility that previously belonged to multiple people. No amount of AI can understand or reconstruct these informal power structures and dependencies. This is why the rehires were necessary: the organization had been wounded in ways that AI couldn’t heal.

A Pattern Across the Industry: From Musk to the Present

This isn’t unique to Block or Klarna. When Elon Musk acquired Twitter in October 2022, he eliminated roughly half the workforce (over 3,000 employees) in early November. Within weeks, he began rehiring dozens of those same people—employees whose dismissal had been hasty, whose expertise proved irreplaceable, or whose roles simply couldn’t be left vacant.

Each wave of rehiring is framed as an exception, a mistake, a correction. In reality, it’s a pattern. Companies announce grand restructurings justified by AI transformation, discover that transformation is incomplete or impossible, and quietly bring people back.

What Jensen Huang Actually Said About Layoffs

During NVIDIA’s GTC 2026 conference, CEO Jensen Huang offered a public rebuke of this trend. He criticized leaders who justified layoffs through AI efficiency gains, stating that such decisions revealed poverty of imagination. “Those leaders who rely on layoffs to cope with AI,” Huang suggested, “simply can’t think of better solutions. They have no new ideas left. Even with the strongest tools, they won’t use them for expansion.” His argument was that AI should enable growth and new business development, not cost reduction through headcount elimination.

Huang’s critique was pointed: if companies truly understood AI’s capabilities, they would be hiring more aggressively, not cutting staff. Layoffs represent not AI-driven transformation but management failure masked by technological justification.

The Uncomfortable Truth: Layoffs as Cost Cutting, Not AI Evolution

Strip away the rhetoric, and the pattern becomes obvious. AI has become the universal excuse for what is fundamentally a cost-reduction exercise. Companies facing stagnant growth, shrinking profits, and business model exhaustion can now blame external technological forces rather than internal strategic failure.

The narrative follows a predictable arc: declare that AI has made certain roles obsolete, announce layoffs as inevitable, cut budgets and headcount, pile remaining work onto remaining staff, and if absolute disaster strikes—if critical institutional knowledge walks out the door—quietly rehire a fraction of those dismissed. The rehires are explained as corrections or exceptions, not as admissions that the initial decision was fundamentally unsound.

This approach is particularly common in Silicon Valley, where layoffs have become a management tool as much as a strategic necessity. When companies cannot sustain their growth narrative through product innovation, they sustain it through cost reduction. AI provides the justification; layoffs provide the immediate result.

The Future Without Magic Cures

AI will eventually change many aspects of work and business. But no technology is powerful enough to compensate for strategic stagnation, obsolete business models, and passive management. AI cannot solve problems created by poor planning, market misjudgment, or organizational incompetence.

The cycle of AI-justified layoffs followed by quiet rehiring reveals something less glorious than technological transformation. It shows that even before the AI era has fully arrived, some people have already been harmed by its anticipation. They’ve been discarded based on promises that companies made to their shareholders and boards, promises that corporate reality couldn’t honor.

For those laid off and then recalled, the experience isn’t a vindication—it’s a scar. They’ve been treated as disposable based on corporate guesswork. And for those watching from their remaining desks, already drowning under the intensified workload of AI-era productivity demands, the message is clear: your job is secure only until the next quarterly earnings report makes it expendable.