I'm pioneering a new way to write blog articles. In this attempt, the base article was drafted by my Second Brain and Opus 4.7, but the examples were written by me. Hopefully, the views will vindicate this new approach to content in the publication creation. In case you are wondering, my AI articles have drastically outperformed my own personal musings, primarily because I direct my AI to write articles relevant to the public that I am too lazy to write myself.
Every few years, the Singapore government rebrands its adult-education machinery, and every few years, a fresh cohort of middle-aged PMETs convince themselves that perhaps studying for an ACLP qualification will save them from the AI tide.
It won't.
And the recent restructuring of SkillsFuture, far from fixing the problem, has merely confirmed it.
The Old Regime: Enrollment as the KPI
When SkillsFuture Singapore (SSG) sat squarely under the Ministry of Education, the dominant performance metric was enrollment. This is what happens when you let educators run an employment programme: you optimise for bums on seats, certificates issued, and modules completed. MOE was very good at this. They are, after all, the same people who ran the most enrollment-obsessed school system in Asia for three decades.
The result was predictable. Singaporeans queued up to spend their $500 (and later, the top-ups) on courses ranging from barista training to wine appreciation to "Introduction to Blockchain." Training providers, sensing a captive subsidised market, churned out catalogues thick enough to double as door-stoppers. Attendance numbers looked fantastic in the parliamentary written replies. Whether anyone actually became more employable was, charitably, a secondary concern.
My own experience with the ACLP qualification was extremely negative; the training materials seemed out of sync with actual industry experience, and there was a lot of negativity in class, as some students were seeking employment with it and rapidly became disappointed that no outplacement services existed.
The early efficacy data tell the story. Around 64% of disrupted workers under 40 found employment within six months of completing a course; only 56% of those over 40 did. Read that the other way: in the cohort most desperate for the scheme to work — older, displaced PMETs — nearly half remained unemployed half a year after their reskilling. (MOE parliamentary reply)
The New Regime: Selection Bias as the KPI
At Budget 2026, PM Lawrence Wong announced that SSG and Workforce Singapore (WSG) would be merged into a new statutory board, jointly overseen by MOE and MOM. (Mothership, Joint MOM-MOE Statement) The official rationale is "tighter integration of the jobs-skills ecosystem." The unofficial rationale is that the old KPI was indefensible.
Notice what happened from 1 January 2026: roughly 9,500 courses across 500 providers were placed under tighter funding criteria. (The Economy) PWM-targeted courses now sit uniformly at Tier 2 funding with employer-sponsorship checks at renewal. Funding favours courses with "direct impact on employability" or "skills demanded by employers." Individual-initiative learning — the experimental, curiosity-driven sort — is being quietly squeezed out.
The new agency will look more impressive on the dashboard. Its outcome metrics will improve. But this is selection bias, not pedagogy. If you only fund the courses whose attendees were already going to be employed (because their employers sponsored them, because they were already on a wage progression track, because they passed a pre-screen), of course, the post-course employment numbers go up. You haven't trained anyone into employability — you have simply screened in the people who would have stayed employable anyway.
The structural critique writes itself: a shift from "education for anyone who wants it" to "education for those most likely to make the numbers look good." MOM brings to the table what MOE never had — labour-market discipline — but it also brings what MOE never wanted: rationing.
I had another personal disappointment lately while trying to get into a fairly new AIxTech training program. While I was not openly rejected by the program, my profession is considered undeserving of a full subsidy. To give you an idea of how different the amounts are, if I were already a software engineer or someone in a list of approved professions, I would need to pay about $200 for the course. As I am a lecturer (in a poly no less, teaching legal technology and data analytics). I would have to 10x the amount, to close to $2,000, to attend the program. Before you get upset at the authorities, they were actually really helpful, and they were asking me to apply under a different profession.
But the fact of the matter is that I don't wish to blow half my Skillsfutures credits to attend an AI course because things are changing so dynamically, so instead I decided to pick up a book on Rust Programming and The Pragmatic Programmer to learn the hard way. But do remember that a great training course is not just about acquiring knowledge but also about building career networks.
What Neither Regime Can Do
Here is the part that both the old enrollment maximisers and the new outcome optimisers refuse to acknowledge publicly. No SkillsFuture course in the current catalogue can transform a worker rendered obsolete by AI into an employable asset.
The reasoning is structural, not motivational. AI displacement is not a "skills gap" of the kind SkillsFuture was designed to address. It is a wholesale revaluation of cognitive labour. A 50-year-old paralegal whose document-review work has been ingested by a large language model does not become employable by completing a 40-hour WSQ module on "Generative AI for Professionals." The module will teach them to type prompts. It will not give them ten years of model-eval intuition, the engineering judgement to ship production systems, or the domain credibility to be hired for AI-adjacent work over a 28-year-old who has been building with these tools since university.
The honest comparison: low-skilled workers are the most exposed to AI displacement, while high-skilled professionals in data science, cybersecurity, and AI development see their roles enhanced, not erased. (NCBI study) SkillsFuture serves the middle of that distribution — the very segment for which short-form reskilling is least likely to bridge the gap.
The Trainer Problem
There is a deeper, less polite reason SkillsFuture cannot solve the AI problem: the trainers will always be behind the curve.
Think about who teaches a SkillsFuture-approved course. They have to be a registered trainer with a Training Provider. The Training Provider needs SSG accreditation. The course needs to be approved against a skills framework. By the time the bureaucracy has finished blessing the curriculum, the underlying technology has shipped two major versions. The frontier of practice — the actual GitHub repos, the Discord servers where the next abstraction is being argued out, the arXiv papers from last Tuesday — never reaches the WSQ syllabus and could not do so within any timeframe that matters.
I say this as someone who teaches for a living. Andragogy — teaching adults — is hard enough when the subject is stable. When the subject changes weekly, the institutional trainer is structurally disadvantaged against the self-directed learner who reads the primary source the day it drops.
What Actually Works
If you've absorbed everything above, the prescription is obvious and unglamorous.
Build a curious mind, and spend more time in the library. That is the entire programme. There is no certificate, no $500 credit, no statutory board to administer it. The people who survive AI disruption will be the ones who treated their own learning as a serious, daily, unsubsidised practice — reading widely, building small things, breaking them, asking better questions than the model can answer.
The library is doing more for your employability than the entire SkillsFuture catalogue. So is a $20-a-month ChatGPT subscription used aggressively. So is sitting down with a textbook on linear algebra at 45 and grinding through it the way you should have at 19. None of these activities is eligible for funding. All of them work.
This is not a counsel of despair. It is a counsel of agency. The state cannot reskill you out of AI disruption because the state's machinery is slower than the disruption. You can. The instrument is your own attention, applied daily, to whatever the frontier currently is.
The Honest Summary
The old SkillsFuture sold us enrollment. The new SkillsFuture will sell us flattering outcome statistics produced by harder pre-screening. Neither delivers what was promised, and neither can. The disruption is structural; the response has to be personal.
Save your credit. Borrow the books. Ask better questions. The library is open.
Sources cited inline; see Joint MOM-MOE statement, Mothership coverage of Budget 2026, MOE parliamentary reply on employment outcomes, 2026 funding tightening, AI displacement and digital skills.
No comments:
Post a Comment