Judgment, proximity, and apprenticeship are pivotal in a market conducive for AI investments, but far less patient with raw potential.
Most advice aimed at young people entering the workforce these days tends to rest on the assumption that the main challenge facing young workers is a skills gap. Some, comparing local graduates with foreigners, go further and lament that locals simply aren’t as hungry.
This is familiar territory for the system – universities, employers, governments – because it places responsibility squarely on the individual and asks institutions to do very little.
But this year’s Singapore Graduate Employment Survey data does not support this narrative:
- Full-time permanent employment for fresh graduates fell from 84.1 percent in 2023 to 74.4 percent in 2025, a 9.7 percentage-point collapse in two years.
- The proportion who applied for jobs and received no offer more than doubled, from 4.1 percent to 8.5 percent.
- Median salaries stagnated at S$4,500, which means the firms that are hiring are paying market rates.
To be sure, the GES figures are annual snapshots. They are not as comprehensive as a proper longitudinal study – something I’ve been calling for since 2018, because I felt it was important for the public to arrive at a more balanced view of local workers who came through the private education pathway.
But directionally, these GES figures still raise an important question contemporaneously: Are the doors to a first job simply not opening, regardless of what is on one’s CV? After all, if skills were the main bottleneck, we would expect to see stronger salary pressure, with employers bidding up pay to compete for scarce talent. Instead, salaries have stagnated.
More telling still, is the doubling of graduates who applied and received nothing. That is not purely a skills mismatch, but reflects a structural demand problem, where fewer firms could be willing to absorb the time, cost, and risk of developing someone from scratch.
Blaming AI is a lazy explanation
The popular explanation for all this is AI: Machines are eating junior jobs.
But the most rigorous research to date, from Peter John Lambert at the London School of Economics and Yannick Schindler at the Ellison Institute of Technology, tells a more complicated story.
They analyzed 243 million new hires and 407 million job postings across the US, UK, Canada, and Australia from 2017 to 2025. When they ran AI exposure and remote work as competing explanations for the junior hiring decline, the AI effect attenuated sharply – in fact, for many specifications it disappeared entirely.
Remote work – not AI – remained the robust predictor throughout. To be clear, the study covers Anglophone labor markets, not Singapore directly. But the mechanism – that remote supervision costs suppress junior hiring – appears consistent with what our own GES data shows directionally.
As John Burn-Murdoch put it in the Financial Times, new evidence suggests the rise of working from home has made entry-level hires a less attractive proposition.
Hiring a junior is fundamentally an apprenticeship bet. In an office, that bet is cheap because seniors are able to mentor these juniors incidentally, juniors absorb these lessons and grow by proximity, and managers can correct their mistakes in real time. In a remote environment, however, every act of supervision becomes a scheduled calendar event. That results in the cost of developing a junior exploding. So rational employers, understandably, raise their bar, hire fewer people who need development, or simply do not backfill.
This matters enormously for the advice we give our young people. If the primary cause is remote work stripping out the apprenticeship infrastructure, then telling a 23-year-old to “learn AI tools” is simply answering the wrong question.
That said, here’s what the AI race really means for you
Here is the global context worth understanding, because it shapes everything downstream.
The United States has committed extraordinary resources to AI – $109.1 billion in private investment in 2024 alone, nearly twelve times China’s $9.3 billion.
The explicit goal of US enterprise AI is to automate knowledge work, compress headcount, and return gains to shareholders. When the major technology firms talk about “productivity”, they really mean fewer people doing more work. This is capital-optimising AI, not worker-empowering AI.
China, on the other hand, is deploying AI into the production function of the world’s largest manufacturing economy.
Industrial AI penetration in Chinese enterprises jumped from 9.6 percent to 47.5 percent in a single year. Smart factory AI cut process development times by 60 percent. China already has 470 industrial robots per 10,000 manufacturing workers. And DeepSeek’s breakthrough – training a frontier model at roughly 5 percent of what OpenAI’s GPT-4 cost – collapsed the cost of AI deployment for every factory floor and SME.
The verdict on the economies for both sides, however, is sobering.
Goldman Sachs found no meaningful relationship between AI adoption and productivity at the economy-wide level in the US through Q4 2025. Only 10 percent of S&P 500 management teams even quantified AI’s productivity impact on their own business.
To be sure, the returns are real but narrowly concentrated at the top of the capital stack. This matters because Singapore’s graduate hiring market is not insulated from these forces. Our knowledge economy is precisely the terrain where US-style enterprise AI lands hardest, and where remote work is most normalized. The average worker – especially the entry-level worker – is experiencing the restructuring costs without the productivity gains.
In Singapore, the IMF estimated that 77 percent of our workforce face high AI exposure – the highest among advanced economies – because our knowledge economy sectors, the ones that dominate graduate hiring here, are precisely the ones most automatable and most compatible with remote work.
Singapore graduates thus face a Lambert-Schindler double jeopardy.
First, Singapore’s MOM has acknowledged higher PMET retrenchment and vacancy rates, characterizing them as AI-driven restructuring rather than broad-based displacement. On the other hand, Budget 2026 debates warned that Singapore risks a “missing middle,” in which AI squeezes junior roles from below, while compressing mid-level execution from above.
Affecting workers in different career phases
In reality, the ‘squeeze’ is not uniform, and understanding where it falls hardest is itself useful, because even if you are just starting out, what happens to the mid-career and senior cohorts above could determine the environment you are entering, and the talent pipeline you are joining.
- If you are 0 to 5 years in a role, the primary challenge right now is the broken apprenticeship model, not simply AI displacement. You need to be near people who will teach you, correct you, and give you real feedback in real time. In reality, the remote work environment has made that informal model of apprenticeship harder to find and harder to justify for employers. That means the firms that are still willing to invest in junior development remain disproportionately the ones worth working for.
- If you are 5 to 15 years into your career, the danger is structural and present-tense. Enterprise AI is compressing mid-level execution – the analyst, the account manager, the coordinator. A senior professional equipped with AI tools can now do the work that previously required a team of associates. This is the zone of quiet displacement: No mass layoffs, just slower backfill and narrower scope.
- If you are a senior employee, the short-term picture is actually better: AI augments judgment and institutional knowledge rather than replaces it. PwC’s 2025 Global AI Jobs Barometer found wage premiums for experienced workers who could demonstrate genuine AI complementarity with domain expertise and fluency. But in the long-run, sustaining this positive outlook depends also on the junior pipeline beneath remaining healthy. Currently, that pipeline appears to be coming under strain.
So, what can 0 to 5s do?
First, engineer your own apprenticeship.
Given that remote work is the primary mechanism suppressing junior development, being deliberate about proximity to experienced people is now a strategic career decision, not a personal preference.
Favor roles and teams where seniors are physically in the room, review your work, and bring you into real decisions. The people who will be genuinely capable in five years are the ones who found environments where they could be stretched and corrected in real time, not the ones who spent three years working from home collecting online badges.
Second, use AI as a force multiplier – for yourself and your team.
Agentic AI tools can now plan, execute, and iterate across complex tasks with minimal human input. For someone early in their career, this is an asymmetric advantage: You can operate above your nominal experience level if you deploy these tools deliberately.
But the individual gain is only half the equation. Firms are under real pressure to justify substantial AI investment, and most are struggling to do so, as I explained in my earlier essay “Singapore’s AI wager: Back the workers, not just the firms”. That’s because the bottleneck is rarely the technology – it is adoption, workflow integration, and the cultural diffusion of AI fluency across teams. In this context, a junior who actively builds that fluency in the people around them, normalising the tools and sharing what works, is solving a firm-level problem. That kind of contribution is visible to leadership in ways that individual output often is not.
Third, hone judgment over speed.
AI and automation are compressing everything that can be made fast and routine. Your sustainable edge is not speed alone, but the ability to understand context, read what is actually happening in a room, push back on a brief that is technically correct but strategically wrong, and take responsibility when something goes wrong. These are parts of the job that are not getting automated soon. They are also the parts no training programme can simply hand you.
As Temus’ CEO Sng Rng Yeong said in his first town hall to our roughly 500-member team: “It’s all about velocity – defined as speed against direction, not speed alone.” Velocity is not just about moving quickly. It is about moving quickly in the right direction. And that kind of judgment compounds only through experience, supervision, and deliberate exposure to decisions that matter.
The advice most often given to young people today – embrace AI, stay curious, build your personal brand – is sensible, but insufficient.
That is because it does not account for the structural reality of a buyer’s market, the broken apprenticeship model, or what has become increasingly scarce in a time of machines: Judgment, proximity, and the willingness to do hard things in the physical presence of people who can actually impart them to you. That, in my view, is the human capital we should be encouraging our young people to build.
Marcus Loh is the Chairman of the Public Affairs Group at the Public Relations and Communications Association (PRCA) Asia Pacific and a Director at Temus, a Singapore AI and digital services firm. Formerly the President of the Institute of Public Relations of Singapore, he helped strengthen the role of strategic communication and public affairs amid shifting policy, technological, and geoeconomic landscapes. He is currently an MA candidate at the War Studies Department of King’s College London.
Temus and PRCA Asia Pacific are pleased to host young participants from Advisory Singapore on 3 June for a candid conversation on first jobs, AI, and what aspiring AI and consulting talents need to build now.
TNGlobal INSIDER publishes contributions relevant to entrepreneurship and innovation. You may submit your own original or published contributions subject to editorial discretion.
Featured image: Anastassia Anufrieva on Unsplash
Convergence in the blood: How AI and biotech are rewriting cancer detection across Asia