Prime Minister Lawrence Wong is right that artificial intelligence can delete the grunt work that once taught graduates how offices actually function. He is wrong if he thinks another awareness campaign will fix it.
At Monday's DBS forum, Wong said employers must not automate away mentorship along with inbox sorting. InfoHandle Editorial agrees — but words at a bank conference are cheap while headcount plans are filed in secret.
Make training measurable
SkillsFuture credits and polytechnic partnerships look good in budget speeches. What is missing is a hard metric: hours of supervised client work per junior hire, reported to the Ministry of Manpower for firms above a revenue threshold. If AI copilots draft the first 80 percent of a memo, managers should log the remaining 20 percent that a trainee still owns.
Stop subsidising hollow internships
Too many "AI transformation" projects use interns as prompt testers without teaching them the underlying business. Tax incentives for digital adoption should require apprenticeship plans that spell out which decisions remain human-only — dispute calls, on-site inspections, patient triage — and who signs off on model output.
Wong hinted at new schemes to nudge employers. We would rather see penalties: if a company takes government AI funding but cuts graduate intake without a workforce plan, claw back the grant.
Workers need bargaining power too
Junior staff should be able to request training records when promotions stall. Unions and professional bodies can publish anonymised benchmarks so a law associate or bank analyst knows whether their firm is an outlier in dumping review work on bots.
The bottom line
Singapore cannot be both the region's AI pilot site and a place where twenty-somethings never learn to negotiate a supplier contract. Wong named the risk; now ministries must audit payrolls and project subsidies against it. Otherwise the bottom rung he fears will stay broken — just with better slide decks explaining why.
What good looks like
We have seen banks pair copilots with mandatory shadowing weeks for new hires; that model should be the baseline, not the exception. Publishing aggregate training hours would let policymakers reward outliers instead of subsidising every firm equally.
Until then, Wong's warning will echo every conference season while graduate employment surveys quietly soften. The city can afford the technology; it cannot afford a generation that never learns to fix what the models get wrong.