When India’s AI turn meets labour realities

As India convenes global leaders to discuss artificial intelligence later this week, the emphasis has largely been on scale and acceleration. Larger models, deeper integration, domestic infrastructure, and sovereign capability dominate the agenda. These ambitions are understandable. Still, they obscure a more fundamental structural question.

In a labour-abundant economy where job creation for a demographically advantageous position of participating youth in the workforce, what does it mean to rapidly scale a technology whose global trajectory has been labour-saving or labour-substituting? AI does not simply increase output. It reorganises production by reallocating tasks between humans and machines. Some activities become automated because they are codifiable, repeatable, and measurable. Others are created because new complementarities emerge between technology and human skill. Whether AI broadens opportunity or compresses it depends on which of these forces dominates.

In advanced economies facing labour shortages and high wages, automation aligns with structural constraints. Substituting capital for labour can raise efficiency without destabilising employment. India’s conditions are different. Millions enter the workforce each year. Informality (nearly 90%) remains the defining feature of employment. The margin of adjustment is not scarcity of labour but of stable jobs.

Much of the formal sector expansion of the past two decades has relied on routine cognitive work in services. Entry-level coding, testing, documentation, customer support, and back-office processing have absorbed graduates from a rapidly expanding higher education system. These are precisely the tasks contemporary AI systems perform most effectively. They excel in environments where outcomes are clearly defined and training data is abundant.

The immediate risk is not wholesale technological unemployment. It is compression. The entry layer of formal employment weakens before new, higher-value tasks emerge at sufficient scale. India’s development path has depended on gradual upward movement from low-productivity work into structured, salaried roles. When the first rung of that ladder narrows, the transition slows. Workers displaced from routine roles do not automatically move into frontier occupations. More often, they move laterally or downward, intensifying competition in crowded segments of the labour market. The effect appears less as open unemployment and more as wage stagnation and thinning pathways of advancement.

Global projections suggest AI’s aggregate productivity gains, at least in the medium term, may be incremental rather than transformative. The distributional consequences, however, may be significant. When machines substitute for routine labour, gains accrue primarily to those who own, finance, and control the technology. In economies where capital ownership is concentrated, and core AI infrastructure remains externally controlled, this dynamic can increase the capital share of income without delivering proportionate wage growth.

For individual firms, substituting labour with scalable AI systems is rational. Lower wage bills improve margins. Faster workflows improve competitiveness. But what is efficient at the firm level does not automatically stabilise the economy.

If substitution outpaces the creation of complementary human tasks, the employment structure thins in the middle. Markets reward cost minimisation. They do not ensure that new roles expand quickly enough to absorb those displaced. This is where the distinction between adoption and adaptation becomes critical. Adoption involves integrating existing AI tools into established workflows. Adaptation requires shaping the direction of technological change so that it complements India’s factor endowments.

The country’s comparative reality is not capital abundance. It is labour abundance combined with uneven skills and limited social protection. Replicating the automation trajectory of labour-scarce economies risks importing a bias towards labour substitution into a context where the social cost of displacement is higher. The more viable strategy is to redirect AI towards easy and hard domains of work. Current systems excel in easy tasks, those that are standardised, codifiable and predictable. But India’s structural advantage lies elsewhere. Much of its labour force operates in hard tasks, environments defined by ambiguity, incomplete information, social coordination, and physical adaptability. Technologies that automate easy tasks in India risk eroding an already exposed employment base. Technologies that augment hard tasks raise productivity without collapsing participation. The distinction determines whether AI substitutes for labour or strengthens it.

A structural choice

Intelligent systems can augment frontline health workers, improve agricultural productivity, reduce leakages in public delivery, and enhance small enterprise performance. The issue is not whether AI should expand. It is whether its expansion is aligned with labour market realities. Technological direction is not predetermined. It responds to incentives, taxation, market concentration, and public investment priorities. When capital investment is favoured and labour markets are fragmented, innovation tilts toward automation.

There is also a generational dimension. For two decades, India’s mobility narrative rested on mass higher education feeding into service sector employment. Routine cognitive roles provided not only income, but entry into formal economic citizenship. If that entry layer contracts before new complementary roles expand at scale, the consequences extend beyond labour statistics.

A demographic dividend depends on absorption. Without credible pathways into stable work, the economic promise attached to education weakens. India does not face a choice between technological ambition and social stability. It faces a design choice about the structure of its growth. A labour-surplus democracy that adopts labour-saving technologies without expanding complementary human tasks risks a fragile equilibrium. Output may rise. Investment may flow. Yet the base of stable, entry-level opportunities may narrow.

The central question is whether India will discipline the economic direction of the advanced models. In a labour-surplus democracy, the market incentive will favour labour-saving deployment. The public policy ecosystem must decide whether that incentive alone should shape the structure of growth. AI can either reinforce the compression of middle-skill opportunities or expand the productivity of labour-intensive sectors. Confusing these trajectories would be costly. An economy can modernise and yet narrow its base of participation. The test of this moment is whether India treats AI as a tool for enlarging capability or allows it to consolidate advantage at the top of the income distribution.

(Deepanshu is a professor and Dean, O P Jindal Global University; Ankur is a research assistant with the Centre for New Economics Studies at the university)