AI substitution pressure in India still outweighs complementary work by half
Work under AI substitution pressure still outweighs complementary work by half in India, with fewer than 8% of the employed in complementary occupations in 2025.
What happened
Chief Economic Advisor V. Anantha Nageswaran released a privately authored white paper, 'AI, Automation and India's Workforce: Who Gains, Who's Exposed, and How Policy Can Shape the Future', written by Vidya Mahambare of the Great Lakes Institute of Management, Vivek Jadhav of the Institute of Management Technology, Ghaziabad, and IAS officer Vishnu Venugopalan. It finds India's AI-exposed employment share has risen modestly since 2012 and the quality of that exposure has improved, but fewer than 8% of employed people were in occupations with a realistic chance of AI complementarity in 2025.
Context
The paper splits exposure into two kinds. Substitution pressure is where the technology can do the task in the worker's place; complementarity is where it raises what the worker can produce. The ratio between them has fallen, which is the improvement — but work under substitution pressure still outweighs complementary work by half. Meanwhile the share of employment in routine work exposed to automation pressure rose by 4.8 percentage points between 2012 and 2025. The mix got better while the routine base grew. In absolute numbers India has 46.3 million workers in AI-complementary occupations, against 52.5 million in the United States and 20.9 million in Russia.
Why it matters
Two distributions matter more here than the national average. Exposure is concentrated among workers early in their careers and declines steadily with age: about 29% of those aged 25 to 34 are in AI-exposed occupations, 10 percentage points above the roughly 19% share among those aged 55 to 64. The cohort with the longest working life ahead of it is the most exposed. The second split is regional. The five southern States average 30.9% of employment in AI-exposed occupations, while Uttar Pradesh, Bihar, Rajasthan, Jharkhand, Chhattisgarh and Madhya Pradesh range from 16.8% to 20%. The paper's own reading is that the southern States lead on the quantity of exposure but not on its quality — a high share of exposed jobs is not the same as a high share of jobs that AI makes more productive.
Impact for India
- Workers aged 25 to 34: at about 29% exposure against roughly 19% for those aged 55 to 64, the youngest working cohort carries the adjustment.
- Southern States: an average 30.9% exposure share means the largest reallocation of tasks, with the paper arguing the quality of that exposure lags its quantity.
- Uttar Pradesh, Bihar, Rajasthan, Jharkhand, Chhattisgarh and Madhya Pradesh: exposure of 16.8% to 20% means less immediate disruption, but also a smaller share of jobs that AI could make more productive.
- Skilling policy: with routine automation-exposed employment up 4.8 percentage points since 2012, the target is moving workers from substitutable tasks into complementary ones.
- Employers: only 46.3 million Indian workers sit in complementary occupations today, which bounds how fast firms can turn AI adoption into measured productivity.
Key data
Concepts
Theory lens
Skill-biased technological change — technology does not displace labour uniformly. It substitutes for routine, codifiable tasks and complements non-routine ones, so its effect on a workforce depends on that workforce's task mix rather than on the technology alone. India's mix is still tilted the wrong way: work under substitution pressure outweighs complementary work by half, and routine automation-exposed employment rose 4.8 percentage points between 2012 and 2025. The improving ratio is real, but it is improving on a widening routine base.
Global context
India has 46.3 million workers in AI-complementary occupations, more than Russia's 20.9 million and not far below the United States' 52.5 million. The level comparison flatters India, because it reflects the size of the workforce rather than its composition. That same 46.3 million is fewer than 8% of India's employed, so the absolute count and the share point in opposite directions — India has many complementary workers and a small complementary share.
Stakeholders
Practice Question — from the story
With reference to the white paper on AI, automation and India’s workforce, consider the following statements:
- Fewer than 8% of employed people in India were in occupations with a realistic chance of AI complementarity in 2025.
- The five southern States have an average employment share of 16.8% in AI-exposed occupations.
Which of the statements given above is/are correct?
Reveal answer & explanation
Answer: 1 only
- Statement 1: Correct. The paper finds that although the share of employment with a realistic chance of AI complementarity has risen, fewer than 8% of employed people were in those occupations in 2025.
- Statement 2: Incorrect. The five southern States average 30.9%. The 16.8% figure is the lower bound of the range for Uttar Pradesh, Bihar, Rajasthan, Jharkhand, Chhattisgarh and Madhya Pradesh, which run from 16.8% to 20%.
Therefore, the correct answer is 1 only.
Concept check — test your understanding
In the economics of automation, what distinguishes a task under substitution pressure from one where the technology is complementary?
Reveal answer & explanation
Answer: B
- The distinction is about what the technology does to the task: replace the worker in it, or make the worker more productive at it. That is why exposure is measured occupation by occupation.
- Option A is a common shortcut but the wrong axis. Routine, codifiable work sits across the middle of the wage distribution, not only at the bottom, so pay is a poor proxy for substitutability.
- Option C confuses exposure with adoption. Both substitution and complementarity are properties of the task content and can be measured before any firm adopts the technology.
- Option D is the lump-of-labour fallacy. Neither category makes a prediction about the aggregate level of employment; they describe how tasks change, not how many jobs exist.