The rush to replace office workers with artificial intelligence has collided with an uncomfortable reality: the technology itself carries a hefty and often overlooked price tag. AMD’s chief information officer, Hasmukh Ranjan, recently calculated that training an employee to lean on AI agents, at roughly $200 a week in token consumption, amounts to $10,000 per person annually. Scaled across 90,000 staff, the bill reaches $900 million—a line item that, as Ranjan noted, simply did not exist a few years ago. The arithmetic challenges the narrative behind nearly 40 per cent of US job cuts announced in May, which employers attributed to AI. Viewed from Silicon Valley, the assumption that swapping humans for algorithms yields instant savings is proving dangerously simplistic.

That cost equation is unfolding against a backdrop of profound labour market bifurcation. A PwC study of more than half a billion job advertisements across six continents identifies a ‘two-track’ market: ‘professionalised’ roles where AI automates routine tasks and elevates human judgement, and ‘democratised’ roles where the technology enables non-experts to perform complex work. Yet economists are increasingly alarmed about a vast, less visible tier of middle-skilled office jobs—customer service representatives, bookkeepers, payroll clerks, HR specialists—that together number in the tens of millions and have long served as pathways to middle-class incomes, particularly for women. In the Gulf, the Emirates Skills platform reports that 95 per cent of new job opportunities are now tied to AI and digital infrastructure, with the share of positions demanding advanced competencies climbing to between 43 and 50 per cent.

Education systems are scrambling to adapt. In Beijing, China has eliminated 12,000 arts, humanities and languages degree programmes—nearly a third of all university offerings—and introduced 10,200 new tech-focused courses such as ‘embodied intelligence’ between 2021 and 2025. The shift mirrors a global trend: India, the UAE, Kazakhstan, Spain and the United Kingdom are all weaving AI into national curricula. In Italy, enrolments at online universities have surged 470 per cent over a decade, with 86 per cent of graduates saying digital institutions better meet technological demands. Meanwhile, corporate giants are both investing and stumbling. Meta launched an AI-powered search mode inside Facebook that one analyst believes could generate $10 billion in annual revenue, yet the company’s chief technology officer, Andrew Bosworth, has admitted in an internal memo that the reorganisation of its Applied AI unit was ‘done badly’, crushing morale and trust. The episode prompted Bosworth to invoke Nvidia’s Jensen Huang: ‘AI won’t take your job, but someone who knows AI will.’

The Gulf states are betting on a more deliberate integration. The UAE has established a Federal Authority for AI and Data, and its Tamkeen 5.0 youth programme recently concluded a hackathon on agentic AI, while the Trends research group launched an integrated AI strategy aligned with the national 2031 vision. Officials in Abu Dhabi frame AI not as a cost-cutting tool but as a means to serve humanity and anticipate crises. The emerging picture suggests that the AI transition will not be a straightforward substitution of labour. The hidden expense of tokens, the erosion of middle-tier career ladders, and the urgent need for reskilling all point to a messy, uneven adjustment. For policymakers from Washington to Dubai, the task is to ensure that the productivity gains of AI do not come at the price of social mobility.