The AI layoffs narrative obscures what’s really happening in Singapore

TEXT START: "For subscribers / The AI layoffs narrative obscures what's really happening in Singapore / A wrong diagnosis could lead to a wrong solution."

THE DISSECTION

This is a corporate consultancy dressed as labor market analysis. The author, Thomas Mathew of Cognizant—an enterprise selling AI services and technology infrastructure—delivers a piece that simultaneously diagnoses a problem and positions himself as the vendor of its cure.

The argument runs: layoffs aren't really about AI replacing workers; they're about uneven capability and training failures. The fix is "building AI fluency" across the workforce through continuous, work-integrated learning. The solution is investment in capability-building infrastructure.

The real product Cognizant is selling here is the consulting engagement that follows from this diagnosis.

THE CORE FALLACY

The article's central error is functional: it conflates the symptom (capability gaps) with the disease (structural displacement) and then recommends treating the symptom while the disease runs its course.

The author correctly observes that AI adoption is uneven, that companies underinvest in training, that a divide is forming between "AI-enabled" and "non-AI-enabled" workers. These observations are accurate. The failure is the conclusion: that solving the training problem prevents the displacement.

The Discontinuity Thesis operates on different mechanics. The relevant question is not whether workers can be trained to use AI tools better. The relevant question is whether the work humans do remains necessary at any price point the employer must pay. The author never asks this question. He assumes the answer is yes, because asking it would be bad for Cognizant's business.

He explicitly writes: "AI isn't replacing workers, but augmenting capabilities and redefining how work is being done."

This is the claim. It is not supported by the evidence he cites. His own evidence—the 1990s automation example—shows the opposite: technology concentrated value at supervisory/management levels while production workers were deskilled and devalued. That's not augmentation. That's displacement with a training-as-hospice-care delay.

HIDDEN ASSUMPTIONS

1. Reskilling can outrun replacement. The author assumes workers can acquire AI fluency faster than AI acquires capability to make those same workers redundant. No evidence supports this. The historical precedent cuts against him: automation consistently displaced faster than reskilling programs closed gaps.

2. Corporate incentives will suddenly prioritize workforce investment. He explicitly admits companies currently underinvest in training because "it's not measurable" and "later often means never." His solution is to argue they should invest more. This is not a policy or technical constraint; it's a fundamental misalignment between capital interests and labor interests that the author acknowledges but refuses to name.

3. Augmentation is the primary trajectory, not replacement. The entire "upskilling" frame assumes human-AI collaboration is the durable endpoint. The DT mechanics argue this is a transitional phase—augmentation holds until AI achieves sufficient capability, at which point augmentation flips to replacement because the employer no longer needs the human in the loop at all.

4. Singapore's "high-skill, high-cost" economy is a defense. The author treats this as a reason to invest in capability-building. The DT lens reads it as a target characteristic: high-skill, high-cost labor markets are precisely where AI achieves cost-performance superiority earliest and most completely. The argument is circular: Singapore should invest in AI capability because it has high-skill workers; high-skill workers are the most immediately displaceable by AI.

5. The training solution benefits the workers the author claims to care about. Who receives the continuous, work-integrated, leadership-modeled training? The author doesn't specify, but the implication is "current employees." The DT mechanics argue the primary displacement target is the current mid-skill cognitive worker—the accountant, the paralegal, the analyst, the copywriter. Training them to use AI tools better makes them more productive while employed and accelerates the timeline for their replacement. The training is the mechanism of the displacement.

SOCIAL FUNCTION

Classification: Corporate transition management / Prestige signaling from the vendor class

This article performs several functions simultaneously:

• It acknowledges legitimate anxieties about AI displacement to appear credible
• It redirects concern from structural displacement (which threatens the workforce) to capability gaps (which the author can position Cognizant to solve)
• It offers comfort to policymakers and corporate leaders that "there's something we can do" that doesn't require confronting capital-labor power imbalances
• It positions Cognizant as a socially responsible technology partner rather than a displacement accelerant

This is the precise genre of content that manages the transition period—delaying political reckoning with mass productive displacement by reframing the crisis as a training and capability problem that can be solved through normal business practice and consultancy engagements.

THE VERDICT

The diagnosis is not just wrong. It's a commercial interest dressed as analysis.

The article correctly identifies that uneven AI adoption will create divides. It incorrectly concludes that closing those divides through training prevents displacement. The DT mechanics state clearly: lag defenses delay, they do not reverse. Training is a lag defense. It may extend the productive relevance of some workers. It cannot preserve the mass employment -> wage -> consumption circuit that the post-WWII economic order requires.

The author has confused the transitional phase with the endpoint. In the transition, AI augments human work. In the endpoint—the Discontinuity—the human is removed from the loop because the task no longer requires them. The training program the author recommends will accelerate the journey from phase one to phase two, because it trains workers to use AI tools that are simultaneously being improved to make those workers unnecessary.

Cognizant's stock in trade is helping companies automate. This article is what that looks like when it writes op-eds.