AI unlikely to cause widespread job losses in India, says Goldman economist

TEXT START: India’s labour force is unlikely to face widespread job losses from artificial intelligence, though some positions in the services sector could be affected, Goldman Sachs Group’s Chief India economist Santanu Sengupta said Friday.

The Dissection

The article is not an analysis of AI’s labor consequences. It is a containment narrative. It takes India’s current employment structure—large numbers of workers in construction, retail, and other physical tasks—and presents that temporary composition as protection against technological displacement.

It also dilutes the central issue by mixing AI with short-term macroeconomic reassurance: resilient demand, vehicle sales, inflation, oil prices, interest rates, and foreign inflows. Those facts do not establish that Indian workers will remain economically necessary. They merely show that the existing system has not yet broken.

The duplicated paragraphs reinforce the article’s function as lightly edited financial-news transmission rather than serious structural examination.

The Core Fallacy

The core error is treating “not currently impacted by AI” as a durable moat.

India’s large workforce is not protection. It is a vast pool of labor whose bargaining power depends on continued demand for human work. Construction and retail may resist immediate AI substitution because robotics, deployment costs, fragmented worksites, and weak infrastructure create lag. Those are implementation barriers, not permanent economic exemptions.

The article also confuses productivity gains with social preservation. Goldman’s projected 0.4-percentage-point productivity increase may raise output while reducing the labor required to produce it. “Productivity benefits outweigh job losses” is a statement about aggregate arithmetic, not about whether displaced workers retain income, status, or productive participation.

The relevant unit is not the occupation but the task stack. Call centers, IT services, finance, education, healthcare administration, retail coordination, logistics, supervision, and eventually parts of construction can be decomposed, automated, and recombined. A worker need not be fired for the labor market to die; hiring can stop, wages can collapse, hours can shrink, and each cohort can enter through a narrower gate.

Hidden Assumptions

• Physical work remains permanently human because current AI systems do not directly perform it.
• A large labor force guarantees sufficient future demand for labor.
• AI adoption can be sequenced without competitive pressure forcing faster deployment.
• Productivity gains will be broadly distributed rather than captured by firms, owners, and capital controllers.
• Workers displaced from services can move smoothly into construction, retail, or other lower-productivity sectors.
• “No widespread layoffs” means no widespread economic obsolescence.
• Aggregate growth automatically converts into mass purchasing power.
• Existing demand resilience will survive falling labor income and rising capital concentration.
• Human institutions can preserve stable human-only domains once AI becomes cheaper and better across cognitive work.
• Ten-year forecasts can be treated as evidence of stability rather than a lag estimate.

The most fraudulent assumption is that gradualism changes the destination. It changes the slope of the collapse and gives institutions time to rename unemployment as informality, underemployment, demographic absorption, or “labor-market transition.”

Social Function

Primary classification: copium and transition management.

Secondary classifications: prestige signaling, ideological anesthetic, and partial truth.

The partial truth is that India will likely experience slower and more uneven displacement than economies with heavier concentrations of exposed cognitive and service work. Physical-task concentration creates a real temporal lag. But the article converts a lag into a verdict of safety, allowing policymakers, investors, and employers to describe structural displacement as a manageable productivity upgrade.

Its practical social function is to prevent recognition of the ownership question. The decisive issue is not whether India produces more output. It is who owns the AI systems, who captures the productivity surplus, and what economic role remains for workers whose labor is no longer necessary. The article does not ask that question because asking it would turn a growth story into a power-distribution autopsy.

The Verdict

This is a delay report disguised as a reassurance report. India is not insulated from AI obsolescence; it is entering the process from a lower-productivity, more labor-intensive base, which creates a longer lag and potentially harsher eventual compression.

The article mistakes physical labor, institutional friction, and gradual adoption for permanent human indispensability. Under the Discontinuity Thesis, India’s mass employment circuit remains intact only while AI cannot economically coordinate and replace enough tasks. Once that threshold is crossed, the country’s enormous workforce becomes an overhang, not an advantage: too many people competing for too little economically necessary labor.

The system is not safe. It is simply not yet forced to admit what it is becoming.