The Productivity Paradox — AI's Efficiency Gains and the Risk of Demand Collapse

The Editorial Argument

Every wave of technological automation in economic history has produced a version of the same anxiety: that machines will take all the jobs. Every time, the anxiety has been wrong — not because machines did not replace human workers in specific tasks, but because the economy adapted, created new categories of work, and sustained aggregate demand. The weavers displaced by power looms became factory workers; the factory workers displaced by industrial robots became service workers; the travel agents displaced by the internet became digital marketers.

The question for Artificial Intelligence is whether this historic pattern holds — or whether AI is fundamentally different in a way that breaks the cycle.

Why AI Might Be Different

1. AI Attacks Cognitive Labour Directly

Previous automation waves targeted routine physical tasks (assembly lines, mining, agriculture). AI, for the first time, targets cognitive and creative tasks — writing, legal research, medical imaging, financial analysis, software coding, customer service, data entry, graphic design. These are precisely the tasks that absorbed workers displaced from physical labour.

If AI displaces cognitive workers at scale, there is no obvious next-higher-order category of work to absorb them. The assumption that “new technology creates new jobs” depends on there being some cognitive task humans do better than machines — an assumption that is weakening with each iteration of large language models.

2. The Speed of Disruption

Industrial automation operated over decades — allowing workers, institutions, and social safety nets to adapt. AI deployment cycles are months to years, not decades. A legal firm that deployed 50 paralegals in 2023 may require 5 in 2027, with AI handling document review, case law research, and contract drafting. The institutions — education systems, retraining programmes, social security — cannot adapt at the same speed.

3. The Demand Collapse Risk

The Keynesian insight is that wages are simultaneously costs (for producers) and demand (for other producers). An economy that eliminates wages through efficiency gains without redistributing productivity gains also eliminates the purchasing power that drives growth. Henry Ford understood this — he paid his workers $5/day (double the market rate in 1914) partly to ensure they could buy his cars.

If AI-driven productivity gains accrue entirely to capital owners while labour income shrinks, the resulting compression of consumer demand could trigger a demand recession that no amount of central bank stimulus can easily reverse — because the problem is structural, not cyclical.

India’s Specific Exposure

India’s development model has relied on services sector employment — particularly IT, BPO, and knowledge process outsourcing — as the primary pathway for educated middle-class employment. These sectors are among the most immediately exposed to AI substitution.

The specific risk:

• India’s IT sector (TCS, Infosys, Wipro, HCL) employs approximately 5 million people directly and 15 million+ indirectly
• AI-assisted coding (GitHub Copilot, Claude Code, GPT-4 Code Interpreter) is already replacing entry-level software development tasks
• BPO and KPO (voice, data processing, annotation) are the lowest-rung jobs most immediately threatened
• India’s domestic manufacturing employment — the alternative absorber — has not grown as planned under PLI schemes

What Policy Can Do

1. AI Dividend / Universal Basic Income. If AI productivity gains are substantial, a portion could be redistributed through a sovereign AI fund — taxing AI-generated revenue and distributing dividends. Universal Basic Income (UBI) pilot data from Madhya Pradesh (2011-12) and recent GiveDirectly experiments suggests cash transfers can support consumption.

2. Education transformation. The National Education Policy 2020 emphasises critical thinking and creativity — skills less susceptible to AI substitution. The bottleneck is implementation: India still has a large share of rote-learning-focused curriculum, especially in government schools.

3. Labour Code social security. The Code on Social Security, 2020 (brought into force November 2025) extends coverage to gig and platform workers — the first legislative recognition that the labour market is changing. Extension to AI-displaced workers would require amendment.

4. R&D incentive for AI in new sectors. India can potentially lead in AI applications in agriculture (AI-led precision farming, crop disease detection), healthcare (AI diagnostics in rural settings), and climate adaptation — sectors where AI creates net employment rather than displacing it.

UPSC Relevance

Mains Keywords: Artificial Intelligence and employment, cognitive automation, demand-side economics, Keynesian economics, Code on Social Security 2020, gig workers, Universal Basic Income (UBI), NEP 2020, India IT sector AI exposure, productivity-demand link, technological disruption

Prelims Facts Corner