The AI squeeze: No juniors, no seniors, more burnout

There has been much hand-wringing of late about how artificial intelligence (AI) has hit young workers. In March, the World Economic Forum reported a 35% fall in entry-level hiring over 18 months.

In Singapore, full-time graduate employment has fallen for three consecutive years, from 87.5% in 2022 to 74.4% in 2025, with the share of graduates receiving zero job offers doubling since 2023.

Meanwhile, another trend is taking shape but has drawn less scrutiny. Experienced workers are walking out earlier. Among Singaporean residents above 60, the proportion neither working nor seeking employment rose from 54.9% to 56.3%. In the US, the trend is more pronounced: labour force participation for those aged 55 and over fell to 37.2% in March 2026, its lowest level in over two decades. Once again, the culprit could be AI.

The generational divide

These trajectories warrant attention. The work that both groups once did does not disappear. It is absorbed by the mid-career workers who remain.

How did it come to this? AI was supposed to make work easier, but prevailing evidence suggests the opposite. An eight-month study by UC Berkeley researchers followed a 200-person US technology company and found that employees who adopted AI tools did not work less. Instead, they took on broader tasks, and extended work well beyond their usual hours.

The researchers identified a self-reinforcing spiral: AI lowers the friction to starting tasks, so workers start more tasks, which expands their workload, which makes them more reliant on AI to keep up. The result was not a productivity revolution, but what the researchers termed “workload creep” — deep fatigue with few, if any, productivity gains.

Then there is the growing suspicion that all this is for nothing. A National Bureau of Economic Research (NBER) survey of nearly 6,000 executives across the US, UK, Germany and Australia found that over 80% of companies report no impact from AI on either employment or productivity. McKinsey dubbed this the “gen AI paradox”: nearly eight in ten companies are using the technology, yet roughly the same proportion report no significant bottom-line impact.

These pressures are experienced differently across generations. A 2025 London School of Economics survey found that while AI usage stands at 83% among Gen Z, it remains at 52% among baby boomers. The adoption gap is real. But it is compounded by employer mistrust.

A survey by Generation, a global employment nonprofit, found that despite 89% of employers acknowledging that older workers perform as well as, if not better than, their younger peers, hiring managers still overwhelmingly prefer candidates under 35 for AI-related roles. ManpowerGroup’s Global Talent Barometer tells a corresponding story: confidence in AI technology collapsed by 35% and 25% among baby boomers and Gen X respectively, while 56% of all workers reported no recent related skills development whatsoever.

Pressured to adopt tools they are neither trained for nor trusted to use, in service of outcomes their own employers cannot demonstrate, the older worker is squeezed from both directions. Many choose early retirement, but to characterise their exodus purely as voluntary overlooks an inconvenient reality: some are not retiring because they want to, but because the cost of keeping up has exceeded their willingness or capacity.

AI has not replaced their roles but made remaining in the workforce feel untenable. This is displacement gussied up as choice. There is a bitter irony in the fact that AI, the most powerful labour-saving technology in history, appears to be delivering not liberation but intensification for those who stay, and an exit ramp only for those wealthy enough to take it.

The technology industry has long treated high employee churn as a feature, not a bug. The assumption was that incoming talent brings compensatory knowledge. But when juniors are not being hired and seniors are walking out, there is no compensatory exchange. Senior employees carry firm-specific knowledge that no system can replicate: client relationships, organisational processes and the contextual judgment that comes from years of experience. In the pursuit of automation, we risk losing all this while failing to develop the next generation.

Consider, too, that companies have overwhelmingly oriented AI toward labour substitution rather than enhancement.

There is, moreover, a distinction that deserves more attention: the difference between AI as labour substitution versus worker enhancement. The prevailing orientation has tilted heavily toward the former, as companies adopt AI to reduce headcount and drive short-term gains. Compounding this, older-age bias in the availability of on-the-job training reinforces rather than bridges the generational divide. Organisational capabilities reside in people. Substitution degrades that asset while enhancement develops it.

The costs of this orientation fall disproportionately on those who remain. Mid-career employees are now truly sandwiched between the AI substitution of entry-level work below them, and the leadership vacuum left by exiting experienced workers above. A study by IESE business school examining 138 million US workers in AI-exposed sectors found that firms reduced junior positions by 4% while increasing mid-level positions by the same amount — even as starting salaries for the latter roles fell by 5.9%. The work is not disappearing; it is being reclassified upward, at lower pay. 

Grit, while battling burnout

This is burnout demographic creep. Exhaustion and disengagement are migrating steadily downward toward younger cohorts who lack the financial means to exit. In Singapore, the consequences are already apparent. Jobstreet’s 2025 Workplace Happiness Index: Singapore found that millennials aged 30-44 were the least satisfied generation at 52%, below both Gen Z and baby boomers; satisfaction with stress levels among this cohort stood at just 31%. These are workers in their 30s and 40s, laden with mortgages, dependent children and ageing parents. They cannot afford to leave and will have to endure. A Duke-NUS Medical School and Institute of Mental Health study estimated that mental health-related productivity losses cost the Singaporean economy S$15.7 billion annually.

Older generations of Singapore workers have been beneficiaries of remarkable economic growth. To be sure, that many can now choose early retirement is itself a testament to that growth. But their exit worsens our old-age support ratio, which has nearly halved over the past decade, from 6.0 in 2014 to 3.5 in 2024, and is projected to fall further to 2.7 by 2030.

The seniors who exit the workforce do not thereby escape the technology they left to avoid. Healthcare, financial and social services are all being redesigned around assumptions of technological fluency. The cost of inclusion simply shifts from employer to state. The Seniors Go Digital programme, which trained over 340,000 seniors in basic mobile literacy at a cost of tens of millions, offers a precedent; the skills required to navigate an AI-mediated world are of a different order entirely.

Considering the societal costs

Singapore’s Budget 2026 has made significant commitments in research funding, workforce transformation, and technological literacy. The emphasis, understandably, is on equipping workers with tools and incentivising firms to integrate them. But these commitments would benefit from a broader scope that includes mitigating the societal and human costs of adoption, not just accelerating adoption itself.

This is not to deny the transformational nature of AI technology. It is an acknowledgement of societal blind spots that have formed in the recent rush to adopt. The costs of a compressed, fragile workforce — in burnout, in lost knowledge, in the public expense of retraining those the private sector chose not to retain — will be borne not just by the companies driving adoption, but more consequently by the remaining workers and their immediate social fabric.

Related: Why young Americans are turning against AI | How AI is rewiring white-collar work in China