AI was supposed to destroy jobs. Where’s the carnage?

In early 2025, some of the most influential figures in artificial intelligence issued stark warnings. Dario Amodei of Anthropic and Sam Altman of OpenAI predicted that AI would soon eliminate half of entry-level white-collar jobs or wipe out entire occupational categories. The message was unambiguous: a wave of technological unemployment was imminent, and the labour market would be reshaped by mass layoffs within months. By mid-2026, that predicted carnage has not arrived. Instead, the global labour market is undergoing a quieter, more complex transformation—one defined not by sweeping job destruction but by consolidation, shifting skill requirements, and a measurable but uneven erosion of specific roles, particularly among younger workers and in certain industries.

Understanding this shift matters because it challenges both the apocalyptic narratives and the complacent dismissals. For executives planning workforce strategies, for policymakers designing labour protections, and for professionals charting careers, the real story is not about a binary outcome of jobs destroyed versus jobs saved. It is about a structural reallocation of tasks, wages, and opportunities that is already underway, with tangible consequences for young professionals and for sectors at the forefront of AI adoption. The data from 2026 reveals a labour market in turbulence—not collapsing, but quietly reordering itself in ways that demand attention.

The Numbers Behind the Shift

The most concrete evidence of AI’s labour market impact comes from sector-specific data. In the United States, the technology and finance industries—long considered the vanguard of AI adoption—are shedding an average of 28,000 jobs per month in 2026 due to AI, according to government data cited by The Straits Times. That is a significant, sustained contraction, yet it represents a fraction of total employment in those sectors. It suggests that AI is not triggering a sudden collapse but a steady erosion, concentrated in roles most susceptible to automation: data entry, routine analysis, basic coding, and certain back-office functions. The monthly figure translates to roughly 336,000 jobs lost over the course of a year, a number large enough to disrupt communities but small enough to avoid the kind of systemic shock that early warnings implied.

The demographic pattern is even more striking. Young professionals aged 22 to 25 in software, finance, and creative industries have experienced a 12.8% drop in employment, according to analysis from OECD.AI. This group is disproportionately affected because entry-level positions often consist of precisely the tasks that AI systems can now perform at scale: drafting standard documents, generating basic code, compiling research summaries, and producing routine design assets. The result is a narrowing of the traditional on-ramp into professional careers, forcing new graduates to compete for fewer, more demanding roles that require judgment, client interaction, or strategic thinking—skills that take years to develop and that entry-level workers have not yet had time to build.

Yet the headline numbers obscure a crucial nuance. Only 3.6% of companies have used AI to fully replace a position, according to a survey by FRS Recruitment. The technology is not, for the most part, eliminating jobs outright. It is changing what jobs require. Nearly half of Irish workers in that survey reported fearing job impact, but the fear is more about transformation than termination—a concern that their current skills will become obsolete, that their workloads will intensify, or that their career paths will narrow, even if their job titles remain unchanged. The gap between the 3.6% replacement rate and the widespread anxiety reflects a labour market in which the threat is real but diffuse, felt more in hiring freezes and shifting expectations than in pink slips.

Consolidation and the New Productivity Bar

The most profound change is happening inside organizations, where AI is enabling smaller teams to deliver dramatically larger outputs. At Bolt.new, an AI-powered coding platform, a three-person team uses an AI agent to achieve output equivalent to what previously required 30 to 40 people, according to reporting by The Guardian. That is not a layoff statistic; it is a productivity leap that fundamentally alters hiring calculus. When a company can achieve the same output with a fraction of the headcount, it does not need to fire existing workers to reduce future hiring. The job destruction is invisible—it shows up as positions that are never created, not positions that are eliminated. A company that once planned to hire ten junior developers may now hire two, and the eight missing roles never appear in any unemployment report.

This dynamic is reshaping employer expectations. According to data from ZipRecruiter, 74% of employers now view AI skills as a strong advantage in candidates, and 13% require them company-wide. The message to workers is unambiguous: proficiency with AI tools is no longer a differentiator; it is becoming a baseline requirement, much like email or spreadsheet literacy in previous decades. Roles are being consolidated as AI handles routine components, leaving workers to manage exceptions, provide oversight, and handle interpersonal or strategic tasks that remain difficult to automate. A customer service representative may now oversee an AI chatbot rather than answer calls directly. A marketing coordinator may review AI-generated campaign drafts instead of writing them from scratch. The job title remains, but the content of the work shifts substantially.

Economists Nicholas Bloom of Stanford and Robert Seamans of NYU describe this phenomenon as market “turbulence.” In their framing, AI changes more jobs than it eliminates. A financial analyst may spend less time building spreadsheets and more time interpreting AI-generated models. A junior developer may shift from writing boilerplate code to reviewing and refining AI-generated code. A paralegal may move from document review to case strategy. This turbulence creates winners and losers, and the transition is not frictionless. Workers who can adapt quickly—who can learn to direct AI tools rather than compete with them—may find their productivity and value rising. Those who cannot, or who work in roles where AI substitutes for their core function, face stagnant wages, reduced hours, or eventual displacement.

The Uneven Geography of Disruption

The impact of AI on employment is not distributed evenly across industries, geographies, or experience levels. The tech and finance sectors, which have both the resources and the incentives to adopt AI aggressively, are experiencing the most visible job losses. Creative industries are also feeling pressure, as generative AI tools handle tasks like copywriting, basic design, and video editing with increasing competence. But other sectors—healthcare, education, skilled trades, and logistics—have seen far less disruption, largely because their work involves physical presence, complex human interaction, or regulatory constraints that limit automation. A nurse, a plumber, or a kindergarten teacher faces a very different risk profile than a data analyst or a junior copywriter.

The generational divide is particularly acute. Older workers with established networks and deep institutional knowledge are often better positioned to adapt, even if their technical skills lag. They have relationships, reputations, and contextual understanding that AI cannot replicate. Younger workers, especially those entering the workforce during this transition, face a double bind: they lack the experience to compete for senior roles, and the entry-level positions that once served as stepping stones are shrinking. The 12.8% employment drop among 22-to-25-year-olds is a warning sign that the traditional career ladder is being shortened at the bottom. New graduates are not just competing with each other; they are competing with AI systems that can perform many entry-level tasks faster and at lower cost.

Policy responses remain fragmented. Some governments are investing in retraining programs and AI literacy initiatives, but there is no coordinated global framework for managing the transition. Labour unions in some countries are pushing for protections around algorithmic management and AI-driven performance evaluation, arguing that workers should have a say in how these tools are deployed. The European Union’s AI Act includes provisions for worker rights, but its implementation is still in early stages, and enforcement mechanisms are untested. For now, the burden of adaptation falls disproportionately on individual workers, who must navigate a shifting landscape with limited institutional support.

What Comes Next

Gartner predicts that AI will create more jobs than it eliminates starting in 2028, a forecast that hinges on the emergence of new roles in AI oversight, data governance, and human-AI collaboration. That timeline, if accurate, suggests a multi-year period of net disruption before the labour market reaches a new equilibrium. The question is how many workers can bridge that gap without falling out of the workforce entirely. A worker displaced from a routine analytical role at age 24 may struggle to retrain for an AI oversight position that requires years of experience and technical depth. The transition period is not a neutral waiting room; it is a time when careers can stall, wages can stagnate, and skills can atrophy.

For professionals, the imperative is clear: adaptation is no longer optional. High-skilled workers must continuously upskill, not just in AI tools but in the capabilities that remain distinctly human—judgment, creativity, empathy, and ethical reasoning. Expanding professional networks and cultivating cross-functional expertise can provide leverage in a market where single-skill roles are increasingly vulnerable. A worker who understands both marketing and data analysis, or both software development and client relations, is harder to replace than one who excels at a single, automatable task. For employers, the challenge is to manage AI adoption without destroying the talent pipeline that future growth will require. Companies that automate entry-level work today may find themselves with no experienced mid-level managers tomorrow. For policymakers, the task is to design interventions that protect the most vulnerable workers—those in automatable roles with limited resources for retraining—without stifling the productivity gains that AI can deliver.

The absence of carnage does not mean the warnings were wrong. It means the process is slower, more diffuse, and more insidious than the most alarmist predictions suggested. Jobs are not disappearing in a single, dramatic wave. They are being hollowed out, redefined, and consolidated. The real story of AI and employment in 2026 is not about a sudden collapse but about a steady, structural shift that is already reshaping who works, how they work, and what they need to know to remain employable. The labour market is not ending; it is changing, and the cost of that change is being borne unevenly by those least equipped to absorb it.