Mr Speaker,
The motion before this House calls for an AI transition that does not leave Singapore’s workers behind. The Prime Minister, the labour chief, the Government as a whole have all said in the past months that this is what they intend. What I want to put to examine this afternoon is whether the policy architecture we have is equal to the commitment we are being asked to affirm.
Mr Speaker, every AI deployment a firm makes is, at root, a choice. The firm can use AI to make its existing workers more capable, more productive, more valuable than they were before. Or it can use AI to do without those workers entirely. The economists’ shorthand for this is augmentation as opposed to automation—augmentation, where AI works alongside the worker; automation, where AI replaces them.
Stanford economist Erik Brynjolfsson, one of the leading academic voices on AI and labour markets, has made a convincing case that in an unaided market—without deliberate policy steering in the other direction—incentives systematically favour automation. Firms find it easier and cheaper to deploy AI to replace workers than to retrain them. The tax code, the labour-market institutions, the cost structure of capital all tilt the playing field. Even though augmentation creates more total value over time—more good jobs, broader prosperity, a fairer distribution of the gains—the default trajectory of an unguided system is automation.
The Government’s chosen direction is augmentation. The motion before us assumes augmentation. The labour chief, in this Chamber yesterday, put the same commitment in his own words: “not AI instead of workers, but AI working for workers.” The philosophical direction is settled across the aisle. The substantive question is whether our policy architecture matches it.
There are three places where our architecture is currently miscalibrated. Three places where, today, the system is permitting automation despite promises to the contrary.
The first: a safety net that pushes workers toward the wrong choice
The labour chief said yesterday that AI “is also reshaping PME jobs in higher-end professions like doctors, lawyers and accountants.” The Prime Minister has said much the same. AI will affect Singapore’s professionals—PMETs who have spent years building specialist careers, and who are now being told the ground beneath those careers is moving.
At his May Day rally last week, the Prime Minister also said: “We may not be able to protect every job. But we will protect every worker.” The question is whether the instrument the Government has chosen—the SkillsFuture Jobseeker Support Scheme—delivers on that promise.
The Prime Minister has called the Jobseeker Support Scheme “the Singapore Way”—a more pragmatic, more Singaporean alternative to the redundancy insurance that is the Workers’ Party’s preferred solution. That reads the Singapore tradition backwards.
Mr Speaker, the labour chief said in this Chamber yesterday that financial support during transition is not welfare—it is an investment in worker outcomes. By that test, the Singapore tradition has long built investments of exactly that kind. CPF, MediShield Life, Medisave—universal, contributory, paid out when life’s major contingencies hit. Each catches every worker because the contingency it insures against can hit every worker. That is the Singapore Way.
The Jobseeker Support Scheme is not built in that tradition. It is a tax-funded grant gated on pre-redundancy income—closer in design to means-tested assistance than to insurance against contingency. As currently configured, it pays up to $6,000 over six months, in tapering monthly instalments—$1,500, $1,250, $1,000, then $750 each for the final three—and is available only to workers who earned $5,000 a month or less.
The labour chief acknowledged in this Chamber yesterday that the ceiling excludes PMEs who face the same displacement risk in the AI era, and proposed raising it closer to the PME median gross income level. If this proposal is adopted, it is movement in the direction that the Workers’ Party has argued for across multiple Parliaments. Mr Ng’s proposal moves the line, the Workers’ Party’s would remove it. That proposal lets more workers into the scheme. It does not change what the scheme does for them.
For those who do qualify under any ceiling, the taper carries its own message. A payment that begins at $1,500 and falls every month is not a floor—it is a countdown. And a countdown pushes a worker to take the first offer, not the right one.
MOM’s own data tells us why that matters. Of retrenched residents in Q4 2025, 43.6% of PMETs had not found new employment within six months—that is the cohort the Jobseeker Support Scheme runs out on. And of those who do find work within six months, roughly four in ten return at lower wages than before. They took what was available, not what their experience was worth. Many of us have experienced how the higher up the career ladder we climb, the longer it takes to find your next role.
Mr Speaker, a PMET pushed by a six-month countdown into a lower-paid job they did not want has experienced exactly the automation outcome the Government’s framework was supposed to prevent—with a small cushion attached for the fall. Raising the ceiling widens the cohort. It does not shorten the countdown.
The Workers’ Party has proposed an instrument built in the actual Singapore tradition. Redundancy Insurance pays 40% of last-drawn salary, for up to six months, with no income ceiling, no tapering mechanism. It is funded by employer-employee contributions, in the same contributory model as CPF, and it covers every worker who pays in—including the professionals the labour chief has identified as most exposed, because the contingency it insures against does not stop at $5,000, $7,600, or at any other ceiling Parliament may set.
The Prime Minister said we will protect every worker. The instrument the Government has chosen does not. The Workers’ Party’s does.
The second: a tax code that is silent at the choice it could be steering
Mr Speaker, when a firm is contemplating a major AI deployment, it stands at a fork. Down one path, it retains its existing workers and retrains them to operate alongside the AI. Down the other, it retrenches, runs leaner, and brings in a smaller, AI-fluent workforce. The first is augmentation. The second is automation. What does our tax code say to the firm at that fork?
The current architecture rewards activity. It rewards capital expenditure on AI. It rewards expenditure on training. These are good things to reward. But what the architecture does not currently do—anywhere—is reward the choice itself. A firm that retrenches its existing workers and trains a smaller set of new hires receives the same fiscal treatment as a firm that retained and retrained its existing workforce. A firm that buys AI to replace workers receives the same fiscal treatment as a firm that buys AI to augment them. The tax code is silent at the fork.
And as Brynjolfsson observed, silence at the fork is not neutrality in consequence. Where the tax code does not actively reward retention, the underlying economics tilts firms toward retrenchment—labour, after all, is the most expensive line on a balance sheet, and labour costs are permanent in a way that one-off training costs are not. An unaided market chooses retrenchment.
The labour chief defended the CTC framework in this Chamber yesterday as the mechanism that ties enterprise transformation to worker progression, and proposed expanding it through the new Tripartite Jobs Council. CTC operates at the project level—for firms that engage with it, with grant funding attached. Expanding its reach scales that grant model. It does not change the broader fiscal architecture every firm operates inside, whether or not it has a CTC. And it is the broader architecture that shapes a CFO’s structural decision-making at the fork. In that architecture, the retention choice remains unrewarded.
In February, in this House, I proposed a Retraining Tax Credit: a deduction available only to firms that can demonstrate they have retrained an existing worker into an AI-augmented role rather than retrenching them. It is the missing conditional piece—a fiscal signal precisely at the fork in the road. The current architecture rewards activity. The Retraining Tax Credit would reward the augmentation choice itself.
The third: no way to verify whether augmentation is actually happening
The fourth limb of this motion affirms that economic progress must remain inclusive. That is a commitment about distribution, not just growth. My colleague, Gerald Giam, proposed a National AI Equity Fund to deliver on that commitment structurally. The instrument I propose today is the diagnostic tool that any redistributive mechanism needs to operate on. Because the third condition for an augmentation strategy to be real is verification.
Mr Speaker, augmentation is, in the end, a testable claim. It makes a prediction: that wages, in the sectors where AI is being deployed alongside workers, will track the productivity gains those workers help to create. If that prediction holds, the framework the Government has adopted is being delivered as advertised. If productivity rises in these sectors but wages do not move with it, then what is being delivered is something other than augmentation, whatever language we use to describe it.
Right now, this Parliament has no systematic way of telling which is occurring.
The Government is investing public money at scale in four National AI Mission sectors: advanced manufacturing, connectivity, finance, and healthcare. Public funds are flowing into these sectors through the CTC grants, the newly-formed Tripartite Jobs Council, the Skills and Workforce Development Agency, and various enterprise transformation programmes. These are appropriate investments. But public investment creates a corresponding public accountability obligation. Where public money goes in, the public has a right to know what is coming out—and to whom.
I am asking for a targeted transparency mechanism: an annual AI Gains Audit, scoped specifically to the four National AI Mission sectors to start, reporting to Parliament on how productivity gains from State-backed AI investment are being distributed between wages and returns to capital. Over time, its scope and coverage can be expanded.
In February, in my Budget speech, I framed this as a distribution question. Today, with this motion before the House asking us to affirm that economic progress must remain inclusive, I propose it again as something more fundamental. The AI Gains Audit is the most direct instrument available to Parliament to test whether the Government’s chosen direction—augmentation—is actually being delivered, or whether something else is occurring under the same name. If the gains are being shared with workers, the audit will say so, and the framework will have the evidence to back its claim. If they are not, we will know—before the gap becomes a chasm, and before this motion becomes a statement of hope rather than of policy.
We cannot manage what we cannot see. And we should not affirm what we cannot test.
Conclusion
Mr Speaker, the choice between augmentation and automation is not made in a single Parliamentary debate. It is made, every day, by the architecture of the schemes we run, the tax code we maintain, and the data we choose to publish or not publish. Whatever this House says today, that architecture will keep making the choice on our behalf.
Right now, our architecture pushes workers toward the first available job rather than the right one. Our tax code says nothing to a firm at the fork between retaining its workers and replacing them. And we have built no mechanism to tell whether the gains from public AI investment are reaching the people in whose name that investment is made.
I support this motion. And I urge the Government to give it the architecture it requires.
Thank you, Mr Speaker.
Footnotes
[¹] Erik Brynjolfsson, “The Turing Trap: The Promise & Peril of Human-Like Artificial Intelligence”, Dædalus, Spring 2022. Brynjolfsson is the Jerry Yang and Akiko Yamazaki Professor at the Stanford Institute for Human-Centered AI and Director of the Stanford Digital Economy Lab. The central finding cited here is that the unaided market produces “socially excessive incentives for innovations that automate human labor and produce weak incentives for technology that augments humans,” and that the corrective requires deliberate policy intervention—including in the tax code.
[²] Ng Chee Meng, parliamentary speech moving the AI motion, 6 May 2026. The NTUC Secretary-General’s speech included quoted passages at the following timestamps: [03:17] PME jobs reshaping (“AI is also reshaping PME jobs in higher-end professions like doctors, lawyers and accountants”); [08:18] augmentation framing (“Not AI instead of workers, but AI working for workers”); [16:32] CTC framework defence (“NTUC pioneered the concept of the Company Training Committee, CTC, so that transformation uplifts businesses and workers’ opportunities together”); [18:58] CTC expansion proposal via the Tripartite Jobs Council; [29:12] rationale for transition support (“It is not welfare. It is an investment in worker outcomes”); [29:37–30:03] proposal to expand JSS coverage (“Adjusting coverage closer to the PME median gross income levels would better reflect the realities of the AI-driven disruption”); [33:54] closing line (“every worker matters”). The same substantive points were made in his earlier May Day Rally address (NTUC), 1 May 2026.
[³] Workforce Singapore, SkillsFuture Jobseeker Support Scheme information brochure (February 2025), available at wsg.gov.sg/home/individuals/jobseeker-support. Monthly payments are: $1,500 (month 1), $1,250 (month 2), $1,000 (month 3), $750 each for months 4–6. Payments are capped at the applicant’s previous monthly income.
[⁴] Ibid. Eligibility requires average monthly income of $5,000 or less over the preceding 12 months. A property Annual Value cap of $31,000 also applies. Applicants must have become unemployed involuntarily—for example through retrenchment, cessation of business, or dismissal due to illness or injury.
[⁵] Ministry of Manpower, Labour Market Report Fourth Quarter 2025 (MOM, March 2026), available at stats.mom.gov.sg. The PMET six-month re-entry rate was 56.4%—implying that 43.6% had not re-entered employment within six months of retrenchment. The all-resident six-month re-entry rate was 57.4%. The re-entry rate has trended downward across recent quarters, reflecting longer job-search periods post-retrenchment.
[⁶] Ibid. Degree holders recorded a six-month re-entry rate of 53.9% in Q4 2025—the lowest among all educational groups, recovering marginally from a recent trough of 51.3%. The twelve-month re-entry rate is more stable (mid-70% to low-80% range), suggesting that degree holders do find re-employment but take longer to do so in the immediate months after retrenchment.
[⁷] MOM Labour Market Reports, 2025 series. Approximately 6 in 10 retrenched workers who re-entered employment did so at similar or higher wages; approximately 4 in 10 returned at lower wages. The MOM reports note that workers retrenched from Financial Services, Information & Communications, and Professional Services—sectors with higher retrenchment volumes—were more likely to achieve full wage recovery, reflecting transferable skills. The wage degradation rate among workers in sectors with less transferable skills is therefore likely higher than the headline average.
[⁸] Workers’ Party, Manifesto 2025. RI parameters: 40% of last-drawn salary for up to six months, capped at 40% of median income, funded by equal employer-employee contributions of 0.1% of monthly salary (approximately $5/month for a median earner, of which the employee pays half).
[⁹] The Enterprise Innovation Scheme (Inland Revenue Authority of Singapore) provides enhanced tax deductions on multiple categories of expenditure relevant to this debate. The training arm provides a 400% deduction on qualifying training expenditure (capped at the first $400,000 per Year of Assessment), applicable to courses funded by SkillsFuture Singapore and aligned with the Skills Framework, for YA 2024 to YA 2028. The AI capital-expenditure arm, announced by PM Lawrence Wong in his Budget 2026 statement, extends the scheme to qualifying AI expenditures (capped at the first $50,000 per YA) for YA 2027 and YA 2028. The argument advanced in this speech is not that one rate is higher than the other—it is that neither arm is conditional on the firm’s choice between retaining and retrenching workers. The Retraining Tax Credit proposed here would be the first conditional fiscal instrument in this part of the tax code.
[¹⁰] Budget 2026, National AI Missions. The four sectors are: (1) advanced manufacturing; (2) connectivity, covering airport and seaport operations and logistics; (3) finance; (4) healthcare.
