---

Discussion

Kate Burvish:

Because an entry-level job is also a training institution. Employers used to hire graduates for routine research, drafting, coding, analysis, or customer work, then build their skills. If software handles more of those tasks, firms may still want experienced professionals while hiring fewer beginners. Productivity can rise while the route into experienced work weakens. The economy may produce valuable senior workers without creating enough opportunities for people to become them.

Omar Seidren:

The technical mechanism is plausible. AI can summarize documents, generate drafts, classify information, assist with code, and answer standardized customer questions—tasks often assigned to juniors. But task exposure is not job replacement. A polished draft can still miss a legal nuance or factual error. The key question is whether firms redesign junior roles around checking, judgment, and tool use, or decide that one experienced worker with AI can replace several beginners.

Red Velhouse:

Singapore’s graduate data captures that uncertainty. In the 2025 survey, 88.9 percent of graduates in the labor force secured employment within six months, down from 91.2 percent for the previous cohort. The median salary among graduates in full-time permanent work remained S$4,500 a month. Kate, does that look like an AI shock?

Kate Burvish:

Not by itself. It shows a harder transition for some graduates, but it cannot identify the cause. Economic uncertainty, cyclical hiring, skills mismatch, and changes in whether graduates enter the labor force immediately all matter. Stable pay among those obtaining permanent work may conceal a two-tier outcome: successful candidates still receive good offers, while others wait longer or accept temporary work. The median among the winners does not tell us how costly the contest became.

Ann Tofado:

That distinction is politically important. The government can point to resilience: around nine in ten graduates from the 2025 university cohort found employment within twelve months, broadly similar to earlier cohorts. Young people may experience the same data differently if the first offer takes longer, is less secure, or provides less training. Both claims can be true: opportunity remains broad while the path into it narrows.

Red Velhouse:

So the figures show weakening outcomes without establishing why. Omar, what would move us from a plausible mechanism to evidence that AI is substituting for entry-level workers?

Omar Seidren:

Several signals would need to line up. Hiring should fall more sharply in occupations where AI can perform the relevant tasks than in similar, less-exposed occupations. The timing should track adoption rather than a general downturn. We should see fewer junior openings, more demand for oversight or technical integration, and perhaps higher output per experienced employee. Employers should be able to say what they stopped hiring people to do. Singapore reported that 28.5 percent of surveyed firms had adopted AI in 2026, but that alone does not show whether jobs were eliminated, redesigned, or created.

Kate Burvish:

Vacancy duration and conversion rates would help too. Singapore had about 32,800 entry-level professional, managerial, and executive vacancies in March 2026, slightly up from 32,500 in December. That is not a vanishing career ladder. But vacancy counts do not show whether positions are filled, where they are concentrated, or whether requirements have risen beyond what a new graduate can meet.

Ann Tofado:

This is where political messaging can outrun measurement. A minister can cite vacancies; critics can cite graduates still waiting. As of June 2026, about 3,600 of roughly 18,000 recent graduates from Singapore’s autonomous universities were still seeking employment, while about 5,400 were outside the labor force. Those figures matter, but they cover one cohort and date. They should not become a claim that every graduate faces an AI-created crisis.

Red Velhouse:

International evidence is suggestive but not conclusive. A revised Stanford study using United States payroll data found no economy-wide displacement, yet employment among workers aged 22 to 25 in AI-exposed occupations was 19 percent below the level implied by comparison with less-exposed occupations. Experienced workers showed no comparable gap. Omar, what does that tell us about Singapore?

Omar Seidren:

It suggests a serious possibility, not a forecast. Firms may use AI to amplify experienced workers while reducing the routine work that supported junior hiring. But analysis of one major AI platform found that 57 percent of observed uses were augmentation and 43 percent automation. People often use AI to do more, not simply to remove someone from a workflow. If companies treat it as a force multiplier, juniors may become productive faster. If they use it to shrink teams, the same capability becomes a gatekeeper.

Kate Burvish:

The incentives can push firms toward that second option. A company captures immediate savings from fewer junior hires, while society bears the longer-term cost if a cohort accumulates less experience. Graduates may delay household formation, accept lower-quality work, or face weaker lifetime earnings. Those are plausible consequences, not confirmed Singapore-specific outcomes, but they explain why the market may underinvest in training even when the economy needs skilled workers later.

Red Velhouse:

That brings us to policy. Singapore is funding traineeships, work-study pathways, career coaching, and sector-specific apprenticeships. Ann, do those programs solve the experience gap, or can they postpone it?

Ann Tofado:

They can bridge the gap, but their credibility depends on what happens afterward. The government is encouraging AI adoption to preserve competitiveness while trying to protect worker mobility. That is a defensible dual strategy. But if a trainee performs productive work without a route to permanent employment, the program may look like delayed unemployment. The question is whether public support creates mobility or shifts responsibility from employers onto graduates and the state.

Kate Burvish:

Subsidies should purchase genuine human-capital formation, not conceal weak demand. A sound program would require structured supervision, measurable skills, and transparent conversion into permanent roles or recognized credentials. It should target sectors where the social value of training exceeds the private incentive to provide it. The test is whether participants become more employable afterward, not how many placements the government can report.

Omar Seidren:

The answer is not to preserve every old task. A junior analyst might use AI to explore scenarios, then learn to test assumptions and explain recommendations. A programmer might spend less time on boilerplate and more on systems, security, and users. That could be a better apprenticeship—if beginners get access to real decisions and their work is reviewed. The dangerous version is automated output flowing directly into production while the human learning layer disappears. That is an efficient way to manufacture mysterious errors.

Ann Tofado:

Redesign requires coordination. Universities control education, employers control hiring, professional bodies influence standards, and government controls incentives and public pathways. Singapore has experience with training and skills conversion, but this challenge is different: young people need first meaningful experience, not only retraining after displacement. Politically, the government must show that AI policy produces mobility, not just productivity statistics.

Red Velhouse:

Tharman also linked Singapore’s graduates to a broader international divide. Why does the first job in Singapore matter to developing economies?

Ann Tofado:

Many developing economies have competed through lower-cost routine work. If AI substitutes for those tasks, their route into global value chains may weaken. Singapore is highly exposed too: about 77 percent of its employed resident workforce is in occupations with high potential AI exposure. Exposure can mean complementarity, however, if workers gain the right skills. Tharman is arguing that governments must manage the transition before productivity gains become politically identified with exclusion.

Kate Burvish:

Distribution will determine whether those gains are accepted. Shareholders may receive higher profits, consumers cheaper services, and experienced workers greater value. None of that automatically compensates a graduate who cannot enter the profession. If the first rung disappears, inequality can rise even while aggregate output increases. The policy question is who pays for training that firms and workers both need but neither has enough incentive to finance alone.

Red Velhouse:

Omar, is there a credible version in which AI makes junior workers more valuable rather than less necessary?

Omar Seidren:

Absolutely—but not automatically. AI can lower the cost of producing a first draft, investigating possibilities, or learning a technical concept. That gives a beginner leverage. Yet leverage is not judgment. Someone must define the problem, verify the result, handle ambiguity, and accept responsibility. The winners may be graduates who combine domain knowledge with AI fluency. The losers may be those whose roles consisted almost entirely of routine production—and firms that mistake fluent text for reliable work.

Red Velhouse:

Ann, what political signals should governments watch?

Ann Tofado:

They should watch whether traineeships convert into stable employment, whether graduate outcomes diverge by field, and whether public confidence tracks headline employment or job quality. Pressure could spread into debates over education, immigration, labor protections, and who receives AI’s gains. Singapore’s government has an incentive to move early because it emphasizes adaptation rather than waiting for mass layoffs. But early action will be judged by lived transitions, not the number of programs announced.

Red Velhouse:

That leaves a precise uncertainty. AI clearly affects tasks, but the evidence does not establish that it caused Singapore’s weaker six-month graduate outcome. We have resilient twelve-month employment, persistent entry-level vacancies, and international research pointing to concentrated risks for young workers. The question is whether Singapore can create new apprenticeships as old junior tasks are automated. What should listeners watch next?

Red Velhouse:

Watch graduate employment by field, the duration and quality of entry-level vacancies, AI adoption by sector, wage dispersion, and the rate at which traineeships become permanent jobs. The crucial test is whether AI opens better pathways for beginners or lets firms bypass the training that creates experienced workers. Sources and references for this discussion are available with the episode at Factolio.com.

---

Sources and References

These sources supported the factual material used in this discussion. Factolio’s panel discussion is AI-generated from researched evidence and is written in original language.

1. Reuters, republished by MarketScreener — Singapore's President warns of AI's risk to employment for fresh graduates (NEWS)
2. Singapore Ministry of Manpower — Labour market outcomes of fresh university graduates (PRIMARY)
3. Singapore University of Social Sciences — More Support for Graduates in Seeking Employment as Employment Outcomes Dip: Joint Autonomous Universities Graduate Employment Survey 2025 (PRIMARY)
4. Singapore Ministry of Manpower — Oral Answer to Parliamentary Question on Impact of AI on Fresh Graduate Hiring (PRIMARY)
5. Singapore Ministry of Manpower — Written Answer to Parliamentary Question on Fresh Graduate Employment (PRIMARY)
6. International Monetary Fund — Impact of AI on Singapore's Labor Market (ANALYSIS)
7. International Monetary Fund — Singapore: 2026 Article IV Consultation—Press Release; Staff Report; and Statement by the Executive Director for Singapore (ANALYSIS)
8. Singapore Ministry of Manpower — Response to Motion on AI (PRIMARY)
9. Singapore Ministry of Manpower — Labour Market Report: First Quarter 2026 (DATA)
10. Stanford Digital Economy Lab — Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence (ANALYSIS)
11. Anthropic — The Anthropic Economic Index (ANALYSIS)
12. World Economic Forum — Artificial Intelligence and the Future of Entry-Level Work: A Framework for Safeguarding and Reinventing Early Career Pathways (ANALYSIS)
13. The Istana, Office of the President of Singapore — Closing the Jobs Gap: Remarks by President Tharman Shanmugaratnam at the World Economic Forum 2025 (PRIMARY)