A Well-Trained Graduate Can Still Be Underemployed

The half that looks like success

Singapore's mid-career reskilling infrastructure is producing strong individual outcomes by the metrics institutions usually report. Close to nine in ten participants aged 51 and above in the Career Conversion Programme remain employed 24 months after completing it. More than six in ten participants aged 51 and above in the Mid-Career Pathways Programme, launched in 2022, found employment within six months of finishing (People Matters, August 2026). By the standard programme scorecard, completion and reemployment, this is a system that is working.

The half that usually isn't measured

The same reporting period carries a second data point that rarely appears next to the first: AI adoption rates by company size. Large companies report 76.4 per cent adoption. Mid-sized firms sit at 54.8 per cent. Small firms are at 27.2 per cent, less than half the mid-market rate and well under a third of the large-enterprise rate.

Put the two figures together and a different picture forms. A programme graduate re-entering the workforce with a newly built skill set is landing, in large part, at employers who have not adopted the technology that skill set increasingly assumes. This is not a flaw in the training. It is a gap in the surrounding system the training feeds into, and it sits entirely outside the metrics most reskilling programmes report on.

Why this matters more as AI fluency becomes the baseline

The gap matters more each year because AI fluency is moving from an add-on to a baseline expectation inside training curricula generally. A graduate trained to work alongside AI tools, placed at a small firm running at roughly a quarter the adoption rate of large enterprises, is functionally underemployed relative to their new skill set, even though every completion and placement metric on the programme's own scorecard reads as a success.

This is the same evidence-versus-outcome gap that shows up across national reskilling systems more broadly. Australia's Jobs and Skills Australia reports that half of advertised positions now require vocational qualifications rather than a bachelor's degree, a genuine expansion of viable pathways. But a pathway that produces a qualified graduate is only half the equation if the employer at the other end of that pathway is not equipped to use what the graduate now knows.

What institutions and funders should be measuring instead

The fix is not more training. It is extending the measurement boundary past the point most programmes currently stop. A completion rate and a 24-month employment rate answer "did this person get trained and get hired." Neither answers "did they land somewhere that can actually use what they learned." That second question requires tracking employer-side readiness, not just learner-side outcomes, and matching graduates against it rather than treating placement as the finish line. In practice that could mean segmenting placement data by employer AI-adoption tier, or building a short employer-readiness check into the placement process itself, so a programme knows not just that a graduate found a job, but what kind of job they found relative to the skill set the programme just built.

For funding bodies designing the next generation of reskilling programmes, that means building an employer-readiness signal into programme design and outcome reporting, not just a learner-outcome signal. Programmes that place graduates disproportionately with small and mid-sized employers should expect, and plan for, a gap between what a graduate can do and what their new employer is ready to use. Ignoring that gap doesn't make it disappear. It just keeps it off the scorecard.