AA Edit: Will AI Be Apocalypse or Game Changer for India?

Artificial Intelligence evokes strong reactions. Its proponents extol the ensuing efficiency-led development. Critics bemoan doomsday job losses. Both are over the top. Job losses in India might be 1.5 million or one half of employment in legacy software firms, as AI agents with reasoning ability, start managing enterprise functions faster, better and cheaper. But new jobs will manage the transition from “dumb” legacy software to AI-driven enterprise solutions. The net job loss will be lower, though individuals who cannot, or do not wish to adapt, shall feel the pain.

AI shall create new jobs. A one-size, one design-fits-all AI plugin across enterprises and domains could potentially wipe out the entire legacy software industry.

Fortunately, this seems unlikely for now. Legacy software systems vary significantly because they adapt to enterprise workflow and functional design patterns. AI solutions will consequently need to align with the individual characteristics of enterprises, creating new jobs for domain AI experts. Software companies are pivoting to fill these transitional roles.

Better, faster, cheaper: Legacy software was faster and cheaper than manual processing of information. Cost reductions ranged between 50 to 90 per cent over two to five years. In comparison, AI offers up to 900 per cent (10X) efficiency improvements by transferring intelligent decision making or reasoning power from humans to chips with enough computing power to make intelligent contextual decisions like the human brain. Net job creation will also happen in producer enterprises and services other than IT, as cost deflation expands demand for new products and services. Early movers in AI integration will also reap the advantages of becoming export competitive.

Race to the bottom for humans: Taking away from humans the unique ability to make complex, reasoned decisions is scary, says Yuval Noah Harrari. But what choice do we have except to follow the emerging trends in machine intelligence.

Mature, self-contained, economically resilient economies like the United States, China and the EU will transition easier as they already have elevated levels of formal employment and stable or ageing populations. The race for getting to the AI frontier is harder for developing economies, because tough choices must be made between outlays on social welfare and investment in technology orientation.

A future of abundance: Elon Musk prophesies a future of abundance when work becomes optional, not a compulsion for humans, because robots would be omnipresent. Even from the perspective of developed economies with full employment, this sounds idyllic. In the developing world where work – especially white-collar work -- is scarce, incomes much lower, and large segments of the population underemployed, the prospect of a meaningful life beyond work is a hard sell.

Funding inefficiency: Developed economies, with abundant fiscal firepower, profligately preserve uncompetitive legacy tasks and jobs. The US government spends about two per cent of its GDP on agriculture, which employs just two per cent of its workforce. The EU spends 0.5 per cent of its GDP on agriculture, which employs about four per cent of its workforce. In comparison, India spends just two per cent of its GDP on agriculture, which employs 45 per cent of the workforce. The tax revenue of the Union and state governments in India is just 19 per cent of GDP, about one half of developed economies, so we cannot afford inefficiency.

Grow, baby, grow: In developing economies, like India, rapid GDP growth via export competitiveness is the key to enhance well-being, build resilience to shocks and provide for national security. Sadly, the growth objective often fails to prevail over preservation of a comfortable status quo, courtesy the near-term compulsions of elective democracy. The low appetite for political risk hobbles higher economic benefits in the medium term from disruptive business opportunities. It does, however, provide political stability, which in turn minimises business risk for incumbents.

The past shapes the future: We are not new to process and employment disruption. During the 1980s and 1990s, the manual processing of high volume, repetitive, rules-based tasks, heavy in information load but low on the need to exercise judgment, were automated. Typing, shorthand, filing, card cataloguing, ledger maintenance and switchboard operations were transferred from humans to machines. In the 2000s, structured transactions like payroll accounting, bank reconciliation, ticketing, directory assistance, basic office workflows followed. Since 2010, cashier services, toll collection, helpdesk, telemarketing and claims processing have similarly been automated.

Data on the employment impact of skill displacement is scarce, even for the “organised” public and private sector, which employs about seven per cent of the total workforce. Public sector jobs in government, and publicly owned banks and corporations declined from 19 million in 1980 to 18 million in 2020. Jobs in the private organised sector (defined as entities with more than ten workers) increased from 5.5 million in 1980 to 13 million by 2020 -- a net gain of 7.5 million.

The public sector socialised the cost of skill redundancy via higher spend so the impact on employees was minimal. Despite the public sector share in GDP halving from 35 per cent in 1980 to 15 per cent in 2025, employment shrunk only by five per cent.

The private sector’s share in GDP increased from 65 per cent in 1980 to 85 per cent in 2025, an increase of 30 per cent. But private sector employment grew by 136 per cent over the same period. This indicates a similar tendency to sacrifice the profitability imperative for socially conscious task transformations.

So, what is different now? First, the speed. Transitions now span two not five years. Technology obsolescence in chips happens within three, not twenty years. Second, the magnitude of skills loss is much higher. Third, the white collar, highly educated class are at risk for the first time, as AI alternatives out compete their reasoning power. Blue collar job loss to robots could happen subsequently, once incomes rise.

What India can do to manage this transition: First, counter intuitively, open the economy aggressively to domestic and imported competition. Suffer now rather than postpone the transition trauma. Inefficiency hides behind tariff walls and moats.

Second, market regulated access to our digital data from a sixth of humanity: a buoyant, native export to keep the current account stable and a bargaining chip for inward FDI in technology. Third, integrate AI for better governance and service delivery in health, education and climate resilience. Finally, end the staged bipolarity between India’s external interests and its domestic priorities. Become the game-changer you want to be.

The writer is Distinguished Fellow, Chintan Research Foundation, and was earlier with the IAS and the World Bank