The number arrived quietly. In its monthly report covering January through May 2026, Challenger, Gray and Christmas, the Chicago-based outplacement firm that has tracked US job cuts since 1993, recorded 87,524 redundancies in which the employer explicitly cited artificial intelligence as the primary cause. No recession. No sector-wide downturn. No financial crisis. The jobs were eliminated because a system could do the work more cheaply, more quickly, and without the liabilities that come with human employment.

The figure was reported, briefly discussed, and then largely absorbed into the ambient noise of the technology news cycle. It should not have been. Eighty-seven thousand redundancies attributed to a single cause in five months is not a data point. It is a signal about the shape of what is coming, and the signal becomes considerably more alarming when you look not at the absolute number but at the trajectory, the sectors affected, and the institutional capacity of economies outside the United States to respond to displacement at anything like this scale.

The Challenger data captures only what employers voluntarily disclose. When a company announces redundancies and states in its press release or SEC filing that the cuts are driven by AI automation, Challenger records that. What it cannot capture is the far larger category of displacement that happens without announcement: the copywriter whose contract was not renewed because the marketing team now uses a large language model, the paralegal whose workload was halved and whose role was reclassified rather than eliminated, the junior analyst whose output was replaced by a dashboard that runs overnight.

Economists studying the broader labour market picture use the term "job task displacement" to describe this phenomenon. A job title persists on the organisational chart while the substantive tasks that justified its existence are progressively automated. The worker remains employed, at reduced hours or reduced seniority, until the role becomes indefensible and is eliminated in the next restructuring round. This process does not appear in layoff statistics. It is nonetheless displacement, and it is happening at a scale that the headline figure significantly understates.

The sectors being affected are worth examining carefully. Financial services, legal services, media and publishing, customer support, and junior software development are not the manufacturing jobs of the previous automation wave. They are knowledge work. They are precisely the category of employment that the previous generation was told to pursue as a hedge against automation. The advice was: do not work with your hands; work with your mind. The current wave of AI is targeting the mind.

Previous waves of technological displacement, the mechanisation of agriculture, the automation of factory assembly, the offshoring of data processing, all unfolded over decades. Economies had time, imperfect and unequal time, but time nonetheless, to redirect labour into new sectors, to build retraining infrastructure, to allow a generation to retire before a successor generation entered a changed labour market.

Generative AI is not operating on a decadal timeline. The capabilities of the systems available in January 2026 are qualitatively different from those available in January 2024. The systems available in January 2028 will be qualitatively different again. The pace of capability improvement is compressing the adjustment window to a point where the institutional responses that societies have historically relied upon, education reform, retraining programmes, social safety net expansion, cannot be designed, funded, and deployed before the next wave of displacement arrives.

In the United States, the federal government spent approximately $3.9 billion on workforce retraining programmes in fiscal year 2025, primarily through the Workforce Innovation and Opportunity Act. The country's ten largest technology companies spent more than that combined on AI research and development in a single quarter of the same year. The investment asymmetry is not a coincidence. It reflects a political economy in which the entities driving displacement have considerably more lobbying power than the institutions tasked with managing its consequences.

The American figure matters for what it signals about trajectory. The more urgent analytical question for a publication focused on the Global South is what happens when that trajectory reaches economies with far thinner social protection systems, far less retraining infrastructure, and far greater dependence on the specific categories of knowledge work that AI is currently targeting.

Three sectors are particularly exposed. The first is business process outsourcing. Countries including India, the Philippines, Kenya, South Africa, and Jamaica have built significant portions of their formal knowledge-economy employment on the provision of back-office services to Western corporations: data entry, customer support, document processing, basic legal research, financial reconciliation. These are precisely the tasks that large language models and robotic process automation systems now perform at a fraction of the cost. The offshoring arbitrage that created those jobs no longer exists when the work can be automated entirely.

The second is content and media work. Across Africa, South and South-East Asia, and Latin America, a generation of educated young workers entered the digital economy as writers, translators, moderators, researchers, and social media managers for platforms and agencies serving global markets. AI-generated content, AI translation, and AI-assisted moderation have compressed the market for this work substantially. Rates have fallen. Contracts have shortened. The entry-level pathways that allowed young graduates to build careers in the digital economy are narrowing.

The third is software development at junior and mid levels. The idea that coding was recession-proof, a skill that would always command a premium regardless of economic conditions, was widely promoted across the Global South as the route to formal sector employment. Junior developers in Nairobi, Lagos, Dhaka, and Ho Chi Minh City are now competing not only with each other and with their counterparts in Eastern Europe, but with AI coding assistants that senior developers use to collapse weeks of junior work into hours.

The standard response to concerns about AI displacement, repeated by technology executives, government advisers, and economic commentators with notable consistency, is that new technologies create new jobs. The industrial revolution destroyed agricultural employment and created factory employment. Computerisation destroyed typing pools and created software engineering. AI will destroy some jobs and create others. Net employment will be broadly maintained. Workers will adapt.

There are several problems with this framing as applied to the current moment. The first is the speed argument already made: adaptation requires time, and the current pace of capability development may not allow it. The second is the nature of the jobs being created. The roles that AI is generating, prompt engineers, AI trainers, model evaluators, safety researchers, infrastructure operators, are concentrated in a small number of wealthy cities, require substantial technical education, and represent a tiny fraction of the displacement volume. The third problem is distributional: even if aggregate employment holds, the displacement falls on workers at the bottom of the income distribution while the gains from AI productivity accrue overwhelmingly to capital owners and highly skilled workers at the top.

The fourth problem is specific to the Global South. The new jobs being created by AI are not being created in Accra or Manila or Dhaka. They are being created in San Francisco, London, and Singapore. The displacement is global. The job creation is geographically concentrated. That asymmetry does not appear in the aggregate figures, but it is the most consequential fact about AI and employment for most of the world's workers.

Eighty-seven thousand jobs in five months in one country. The number will be larger by the end of the year. It will be larger still in 2027. The question is not whether the displacement will continue. It will. The question is whether the political, institutional, and educational systems of countries that have the most to lose from AI-driven labour market disruption will respond with the seriousness and the speed that the situation demands.

At present, most are not. The policy conversation in most developing economies remains focused on AI as an opportunity, the efficiency gains, the productivity improvements, the leapfrogging of infrastructure gaps. Those gains are real and worth pursuing. But the opportunity framing is incomplete without an equally serious engagement with the displacement risk, and the displacement risk is not five or ten years away. For workers in business process outsourcing, content creation, and junior software development across the Global South, it is already present. The 87,000 is not a warning about the future. It is a measurement of a process that is already under way.