Singapore: output up, hiring flat

Singapore's growth forecast for 2026 was raised and the second quarter came in at 5.9 per cent, yet resident employment grew by just 2,200 and retrenchments rose to 4,620, the highest since late 2020. Fewer laid-off residents were back at work within six months. The Ministry of Manpower's own figures explain why this is not, so far, a story about AI replacing people: most adopting firms report better productivity, and only a small slice cut headcount. Our view is that growth arriving through software and compute rather than payroll is the pattern to plan for — and the honest gap is that nobody yet knows whether it is permanent.

The Ministry of Manpower's quarterly Labour Market Report, released on 21 September 2026, put retrenchments at 4,620 for the three months from March to June, the highest quarterly figure since late 2020. Job vacancies fell, and total employment grew mainly on foreign workers in construction and manufacturing, while resident employment rose by only 2,200. Unemployment stayed low, and the number that moved was re-entry: fewer retrenched residents found work within six months than at the end of March. We would watch the re-entry rate rather than the headline, because it is where a soft market shows up first.

The Ministry of Trade and Industry upgraded its 2026 forecast, citing an acceleration in global AI-related capital expenditure. That is the shape of the gap: the growth is being booked in software and compute, and the payroll line has stopped moving in step. One regional economist told The Business Times that the economy should be generating far more jobs a quarter than it is, and warned that the gains may stay concentrated if the AI boom does not translate into broader employment growth and wage gains. We would not read this as AI destroying jobs; we would read it as capital doing what capital does.

The Ministry of Manpower's report on AI adoption among firms surveyed private-sector establishments across Singapore. It found most had not adopted AI at all, and of those that had started, only a small fraction had worked it into core processes, with the rest planning or piloting. Among adopters, most reported improved worker productivity and some redesigned roles, against a small slice that reduced headcount. The constraints were cost and skills rather than ambition. Our reading is that AI is changing tasks rather than cutting roles — so far, and that 'so far' is carrying a lot of weight. That is the part we would underline: the tool is reaching the task, not the person, and a task is far easier to redesign than a role is to replace.

A study from Konstanz, reported on 18 September 2026, found use of AI at work rose only slightly over a year, and that adoption is often improvised rather than managed: only about half of AI users say the tool they use most was introduced by their employer. The divide follows the work more than willingness, thinning sharply in production and manual roles and among the least educated. Employees are moving faster than the organisations they work in. We would rather build a habit than write a policy. The gap between what employers plan and what their staff actually do is the quiet risk running through all of it.

Two lessons survive the contradictions. The productivity gain is real where AI meets a defined process: one Singapore software firm multiplied its releases by reviewing AI's first draft instead of writing it. The second is that the freed hour is worth nothing until someone has decided what it is for. AI-related job postings in Singapore have grown sharply, which tells us the work is moving — and the people who move with it are those who can show what their review of AI output is worth.

We would not read this as AI destroying jobs in Singapore. We would read it as capital doing what capital does: growth booked in software and compute, with the payroll line no longer moving in step. The honest gap in the data is that nobody yet knows whether the missing vacancies are a permanent change in how output is produced or only a cautious few quarters — the ministry itself calls the outlook positive but very cautious.

For a studio our size the practical reading is unglamorous. Decide which hour you are trying to free, then be able to show what filled it, because the firms that will hire next year are the ones that can point at something AI made better rather than something it made faster.