Why so many people fear AI at work: risks, layoffs, and job security

The overlap between AI capability and real work tasks is not speculative: it is now measured, mapped, and visible in the numbers.

A new OECD meta-study published just days ago makes it clear that the apprehension so widely felt is not the result of media panic or baseless rumor, but instead arises from a cold, statistical reality: nearly 40% of global jobs today—across every sector, but particularly in advanced economies where the content of work is more cognitive than manual—now involve tasks that generative AI can already perform at a high level. And the percentage rises to about 60% in the economies where office-based, information-driven work is the norm and the threshold for routine is defined by what can be done with data, text, or code. This is not simply a matter of warehouse robots replacing lift drivers; it is a wave of substitution for the very tasks that used to define professional “knowledge work.”

The World Economic Forum’s Future of Jobs 2025 survey, which collects projections from over a thousand multinationals, now expects that automation, and particularly generative models, will not only create jobs—around 14% of today’s employment is forecast to come from roles that barely exist yet—but will also displace the equivalent of 8% of the current workforce (nearly 92 million jobs) between 2025 and 2030. The most significant transformation, however, is not just in headcount but in the content of roles: WEF expects that 39% of all skills in every job will be outdated within this window, requiring mass reskilling on a scale and with a speed that is unprecedented in labor history. It is these hard numbers, and not simply headlines, that make anxiety rational and pervasive.

Real companies are translating forecasts into payroll cuts, not just process diagrams.

It is not a theoretical risk when multinationals and tech-forward firms begin making structural changes to their teams, guided explicitly by AI adoption. The Dutch navigation firm TomTom eliminated 300 employees on June 30, 2025, folding an entire “application layer” team into a much smaller, AI-centric product group, demonstrating that entire organizational layers can become redundant when new tooling makes old workflows obsolete.

Klarna—a prominent fintech platform—made international headlines by loudly advertising that its GPT-powered assistant “now does the work of 700 agents.” While a month later the company admitted that customer experience metrics declined and it began rehiring human support for specific cases, the net effect is still a 40% reduction in team size, with no intent to return to pre-AI staffing levels.

At IBM, the CEO confirmed in May that “several hundred” roles in human resources—traditionally considered protected, complex, and high-trust—have already been handed to AI systems, with freed budgets redirected to core engineering. The message to every office worker is stark: if the role’s main activities involve repeatable decisions, standardized documents, or structured communication, it is now exposed.

These are not pilot projects or minor tweaks—they are visible, irreversible changes to the corporate headcount, and their example resonates through the entire labor market.

Generative models do not just automate tasks; they erode the “task-core” of many established professions.

Whereas the previous generations of workplace automation targeted physical activity—on the assembly line, in warehouses, in logistics—today’s leading AI models are directly attacking the core activities of traditional white-collar employment: parsing and generating language, drafting or analyzing code, producing reports and presentations, summarizing contracts, and even assisting in design or research.

The latest McKinsey workplace analysis concludes that up to 30% of the average knowledge worker’s week—activities such as building first-draft slides, writing emails, reconciling figures, or basic customer outreach—can already be offloaded to today’s foundation models. And in organizations that move fastest, this does not simply create time for workers to “move up the value chain”; instead, it allows companies to consolidate entire workflows, retaining a smaller group of higher-skill staff while disbanding teams whose primary function has become automatable. The process is gradual and often invisible, happening task by task rather than as a single dramatic layoff, which means that many employees only realize the vulnerability of their job once the very last manual or creative task is automated away.

Workers are not just guessing: they feel, with acute specificity, that AI will erode their prospects and stability.

A February Pew survey of U.S. employees shows the depth of concern with new clarity: 52% of workers now say they are “worried” about AI’s impact on their own work life, and 33% feel “overwhelmed” rather than optimistic or excited. Critically, only 36% report feeling “hopeful.” Among clerical staff, accounting teams, and junior analysts—the precise segments that both the WEF and McKinsey flag as most exposed—the sense of threat is even sharper, with a full 32% of respondents already expecting personally “fewer opportunities” as a result of AI in their sector.

This is not a vague or general concern: it is rooted in the experience of seeing colleagues redeployed, laid off, or reassigned as AI-driven process redesigns trickle through organizations in real time.

The bottleneck in reskilling: a structural gap that will leave millions exposed.

While every optimistic projection from international agencies now insists that reskilling is the answer, the data reveal a grim gap between aspiration and execution. The WEF’s own numbers show that 59% of workers worldwide will require significant training by 2030 to keep pace with changing roles, but that employers themselves believe at least 11% of those affected will receive no reskilling at all—either because programs do not exist, or because the economic calculus favors replacement over retraining.

OECD analysis indicates that today, only 1% of roles are fully automatable, but as organizations move beyond simply adding AI tools “on top” and start to reengineer entire workflows, exposure spikes: more and more jobs will require workers to reinvent their skills before—not after—they become redundant. This demands a timeline for change that is not always matched by corporate budgets or by the willingness of individuals to invest in learning new, often unrelated, abilities at mid-career.

Inequality is poised to worsen: AI’s wage impact runs in two directions, not one.

It is not just the number of jobs at stake, but the distribution of rewards and security. An IMF working paper on the global diffusion of AI finds that, at current rates, early adoption disproportionately increases earnings for those with high-skill, AI-complementary roles, while compressing or eroding wages for those in mid-tier, routine jobs, and doing little to assist the lowest-skilled. The net effect is to widen the income gap (as measured by the Gini coefficient), unless tax, welfare, and training policies intervene. So even in advanced economies where overall employment may not decline, the sense of relative insecurity grows sharper: not only might your job go, but the rungs of the social ladder may get further apart.

Institutions are slow, but technology moves on its own curve: the “asymmetric risk” of delayed regulation and support.

Whereas corporate adoption of generative AI can be measured in quarters—firms deploy cloud-based LLM APIs and change workflows in months—statutory severance, reskilling subsidies, and collective-bargaining agreements typically take years to adapt. The lag between a CTO’s procurement and a parliament’s policy cycle is now a central source of fear: workers face what sociologists call “asymmetric risk,” where the forces that threaten job security act faster than the forces meant to provide stability, fairness, or second chances. For every worker who loses a job because of AI-driven restructuring, there are thousands more who see the headlines and realize there is no quick path back to stability.

The psychological dimension: identity and status are harder to reinvent than skill-sets.

Economic models often count only dollars or roles, but surveys and qualitative research show that for many employees—particularly those with years or decades invested in a profession—what is most devastating is not simply a change in daily activity, but the sudden sense that their skills, training, and status are obsolete, exchangeable, or devalued. As news cycles publicize the replacement of engineers, marketers, or even creative professionals by AI, the implicit promise that higher education and professional commitment guarantee long-term stability is eroded. The loss is not just financial, but existential: work has always been a source of purpose, connection, and recognition. That foundation now seems provisionally, and perhaps permanently, unstable.

Data and detail underpin the fear: measurable exposure, real-world layoffs, and institutional lag combine to make anxiety rational.

The current wave of AI-driven change is not a story of distant science fiction or panic-stoking headlines. It is a process that can now be tracked in data—OECD, WEF, Pew, McKinsey, IMF—and in the decisions and public statements of major employers across continents. Millions of jobs will change, millions will be replaced, and many millions more will be exposed to new forms of uncertainty as both technology and society adapt, at mismatched speeds, to a world where “work” itself is being reinvented from the ground up.

______

FOLLOW US FOR MORE.

DATA STUDIOS

datastudios.org