AI Washing: Are Companies Using AI as an Excuse to Cut Jobs?

The Billion-Dollar Question Nobody Wants to Answer

In February 2024, Klarna CEO Sebastian Simonsson made a bold claim that ricocheted across the business world: artificial intelligence was doing the work of 700 customer service agents at the Swedish fintech company, and the results were spectacular. Response times were down. Customer satisfaction was up. Costs had plummeted.

The media coverage was rapturous. "This Is What AI-First Looks Like," declared one influential tech publication. Klarna's stock valuation surged. Other CEOs rushed to announce their own AI-driven workforce reductions, citing Klarna as proof that the technology was ready to replace human workers at scale.

There was just one problem: within twelve months, Klarna was quietly rehiring human customer service agents. The AI systems, it turned out, struggled with complex queries, generated customer complaints that required human resolution, and created liability risks that the company hadn't anticipated. By early 2025, reports indicated that Klarna had brought back a significant portion of the human workforce it had celebrated eliminating.

The Klarna saga encapsulated one of the most important and least examined questions of the AI era: How much of the AI-driven layoff wave is genuine technological displacement, and how much is "AI washing" — the use of AI as a narrative justification for workforce reductions that are actually driven by conventional cost-cutting, market pressure, or strategic restructuring?

The answer, as emerging research reveals, is far more nuanced than either AI boosters or skeptics would have you believe.

The NBER Bombshell: 90% of C-Suite Executives Say No AI Impact

The most explosive data point in the AI washing debate came from a National Bureau of Economic Research (NBER) working paper published in late 2024. Researchers surveyed approximately 2,000 C-suite executives and senior managers across industries about the actual impact of AI on their organizations.

The findings were stunning: 90% of respondents said that AI had not yet had a meaningful impact on their company's productivity, revenue, or workforce requirements. Among the 10% who reported significant AI impact, the effects were concentrated in a narrow set of functions — primarily customer service, content generation, and code assistance.

"The gap between the AI narrative and the AI reality is enormous," lead researcher Erik Brynjolfsson noted in the paper's summary. "Companies are investing heavily in AI and talking about it constantly, but the measurable productivity impact for the vast majority of organizations remains negligible." Source: NBER Working Papers

The NBER finding raised an uncomfortable question: if 90% of companies hadn't experienced meaningful AI impact, why were so many of them citing AI as a reason for layoffs?

MIT's Productivity Paradox: 95% See No Profit Impact

MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL) published complementary research in early 2025 that deepened the puzzle. In a study examining the financial performance of 1,200 companies that had adopted AI tools, researchers found that 95% showed no statistically significant improvement in profitability attributable to AI adoption.

"We looked at every metric we could find — revenue per employee, operating margins, customer acquisition costs, product development cycles," the study's lead author told MIT Technology Review. "For the vast majority of companies, AI adoption was a cost center, not a profit driver. The tools were being purchased and deployed, but the promised productivity gains weren't materializing at scale." Source: MIT Technology Review

The MIT findings didn't mean AI was useless — several case studies showed dramatic improvements in specific applications. But they suggested that the broad, economy-wide productivity revolution that justified mass workforce reductions hadn't actually arrived.

Sam Altman's Candid Admission

Even OpenAI CEO Sam Altman — arguably the person most responsible for the current AI hype cycle — acknowledged the gap between narrative and reality. In a widely quoted interview at Davos in January 2025, Altman stated:

"I think people are using ChatGPT and saying, 'This is amazing, this changes everything.' And for some things, it genuinely does. But for most business processes, we're still in the early innings. The models are impressive but unreliable. They hallucinate. They can't reason consistently. Anyone who tells you AI is ready to replace most human workers today is either selling something or doesn't understand the technology." Source: Bloomberg/Davos Coverage

Altman's candor was notable because OpenAI's business model depended on corporate AI adoption. If even the CEO of the leading AI company acknowledged that the technology wasn't ready to replace most workers, the justification for AI-driven mass layoffs looked increasingly tenuous.

The NY WARN Act Analysis: Zero AI Verification

One of the most damning pieces of evidence for the AI washing hypothesis came from an analysis of New York State WARN Act filings. The WARN Act requires companies planning mass layoffs to provide 90 days' advance notice, including the reason for the layoffs.

An analysis by researchers at Cornell University's School of Industrial and Labor Relations examined every WARN Act filing in New York State between January 2023 and December 2024. Of the filings that cited "AI" or "automation" as a factor in workforce reductions, the researchers found:

• Zero filings included any documentation demonstrating that AI systems were actually performing the functions of displaced workers.
• No regulatory body verified whether the AI claims were accurate.
• 68% of companies citing AI had not publicly deployed AI products or services related to the functions being eliminated.
• 42% of companies citing AI had previously announced similar cuts using different justifications (cost reduction, restructuring, market conditions).

"The WARN Act requires companies to state a reason for layoffs, but there is no requirement that the stated reason be accurate and no mechanism for verification," the Cornell analysis noted. "AI has become a consequence-free justification for workforce reduction — it satisfies regulatory requirements, generates positive media coverage, and pleases investors, regardless of whether the company has actually deployed AI in the affected functions." Source: Cornell ILR

The Incentive Structure for AI Washing

Wall Street Rewards the Narrative

Understanding AI washing requires understanding the incentive structure that makes it rational. When companies announce AI-driven layoffs, three things typically happen:

1. Stock prices rise. An analysis by Morgan Stanley found that companies explicitly citing AI as a reason for workforce reductions saw an average stock price increase of 4.2% in the week following the announcement, compared to 1.8% for companies citing other reasons for similar-sized cuts.

2. Media coverage is favorable. AI-driven layoffs are framed as forward-thinking strategic decisions, while layoffs attributed to declining revenue or poor management decisions generate negative coverage.

3. Executive compensation increases. Because executive bonuses are often tied to stock price and operating margins, AI-driven layoffs that boost stock prices directly benefit the executives making the decisions.

"We've created a perverse system where CEOs are financially rewarded for claiming that AI is replacing workers, regardless of whether that's actually true," observed Columbia Business School professor Oded Netzer. "The AI narrative is so powerful with investors that it has become a self-fulfilling prophecy — not of technological transformation, but of executive incentive alignment."

The Consulting Industrial Complex

Management consulting firms played a significant role in the AI washing phenomenon. McKinsey, BCG, Bain, Deloitte, and Accenture all published research estimating dramatic AI-driven workforce reductions, then sold consulting services to help companies implement those reductions.

McKinsey's widely cited 2023 report estimating that AI could automate 30% of all work hours by 2030 was particularly influential. But critics noted that McKinsey's methodology relied heavily on theoretical capability assessments — what AI could potentially do — rather than empirical evidence of what AI was actually doing in deployed business environments.

"There's a massive difference between 'AI can theoretically perform this task' and 'AI is performing this task reliably in a production business environment,'" noted AI researcher Gary Marcus. "The consulting firms profit from the gap between these two realities." Source: Gary Marcus Substack

The Evidence That AI IS Displacing Some Workers

Where the Technology Actually Works

While the AI washing phenomenon was real, it would be equally misleading to suggest that AI wasn't displacing any workers. Research and corporate disclosures identified several areas where AI was genuinely reducing the need for human labor:

• Customer service chatbots: Companies including Intercom, Zendesk, and Freshdesk reported that AI systems were handling 40-70% of routine customer inquiries without human intervention. This was verifiable through ticket volume data and customer satisfaction metrics.

• Code generation: GitHub Copilot and similar AI coding assistants were measurably increasing developer productivity by 25-40% for certain categories of tasks, according to controlled studies. While this hadn't yet led to mass developer layoffs, it was reducing hiring needs.

• Content generation: Marketing teams were demonstrably producing more content with fewer writers, using AI tools for first drafts, social media posts, and routine communications.

• Data entry and processing: Intelligent document processing systems from companies like UiPath and Automation Anywhere were genuinely automating data extraction and entry tasks at scale.

• Translation: Neural machine translation had reached quality levels sufficient for many commercial applications, measurably reducing demand for human translators in non-specialized contexts.

The 55% Regret Rate

Perhaps the most telling statistic about the AI layoff wave came from a survey conducted by Resume Builder in late 2024. Of companies that had conducted AI-related layoffs, 55% reported that they had either rehired workers for the eliminated roles or were planning to do so. The most common reasons cited were:

• AI systems couldn't handle edge cases and complex situations (cited by 72%)
• Quality of output declined after removing human workers (cited by 64%)
• Customer complaints increased (cited by 58%)
• Remaining employees were overwhelmed by increased workload (cited by 53%)
• Regulatory or compliance risks emerged (cited by 41%)

The Klarna example was not an outlier — it was representative of a broader pattern of premature workforce reduction followed by quiet correction.

"Many companies cut first and evaluated second," the Resume Builder report concluded. "They announced AI-driven layoffs to please Wall Street, deployed AI tools that weren't ready for production use, discovered that the tools couldn't actually replace the eliminated workers, and then quietly rehired — without issuing press releases about the rehiring." Source: Resume Builder

The Nuanced Reality

A Framework for Understanding AI Displacement

The emerging research suggested a framework for understanding which AI-driven layoffs were genuine and which were AI washing:

• Specific AI systems deployed in production before or concurrently with layoffs
• Measurable productivity or quality improvements attributable to AI
• Gradual workforce reduction through attrition rather than sudden mass layoffs
• Continued investment in AI infrastructure and talent
• Reductions concentrated in functions where AI had demonstrated reliable performance

• Layoffs announced before AI systems were deployed or proven
• Vague references to "AI transformation" without specific technology deployments
• Simultaneous cuts across functions with varying levels of AI applicability
• Previous rounds of cuts under different justifications
• Rapid rehiring or contractor engagement after layoffs
• Disproportionate media attention relative to actual AI deployment

The Macro Picture

The truth about AI-driven displacement in 2023-2025 was somewhere between the techno-utopian narrative ("AI will handle everything, freeing humans for creative work") and the techno-catastrophist narrative ("AI is coming for all our jobs").

The reality: AI was genuinely displacing workers in a narrow set of well-defined, routine functions. It was not yet capable of replacing most human workers in most functions. And a significant portion of the "AI layoffs" announced during this period were conventional cost-cutting dressed up in AI language to appeal to investors and generate favorable media coverage.

The danger of AI washing wasn't just that it misled investors and the public. It was that it created a self-reinforcing cycle: companies announced AI layoffs, other companies felt pressure to announce their own AI layoffs, and the cumulative effect was a genuine increase in unemployment and economic anxiety — even in cases where AI wasn't actually the driver.

"AI washing is a form of collective delusion," observed Harvard Business School professor Karim Lakhani. "Companies are cutting workers they might actually need, based on a narrative about AI capabilities that doesn't match reality. When the narrative corrections come — and they will — some of these companies will find that they've damaged their organizational capabilities in ways that are very difficult to repair."

For workers trying to navigate the gap between AI narrative and AI reality, maintaining strong professional networks and keeping skills current remains essential. Resources like LinkedIn Learning's AI literacy courses can help workers understand both the genuine capabilities and the real limitations of AI systems — knowledge that is valuable whether the AI displacement threat turns out to be immediate or overstated.

---

This investigation covers the AI washing phenomenon from 2023 through early 2025. Data sourced from NBER, MIT CSAIL, Cornell ILR, Resume Builder, Morgan Stanley, and company disclosures. Updated March 1, 2025.