India’s information technology and business process management sector — which generated US$254 billion in revenue and employed 5.4 million people in FY2023–24, according to NASSCOM‘s 2024 Strategic Review — is confronting a structural threat from AI automation that strikes directly at its core proposition: selling human cognitive labour at scale to Western corporations. TCS shed a net 13,249 employees in FY2023–24 while revenues continued to grow, and Infosys recorded its largest annual headcount decline in its history, a 5.9% year-on-year fall to 317,240 staff as of March 2024.
Neither company attributes the cuts solely to AI, but the pattern is unmistakable: output rising, headcount falling. For the millions of Indian engineering graduates whose career expectations were built on a desk in Bengaluru or Hyderabad, that arithmetic is the real story.
The buried threat to India’s economy is not that AI will destroy its IT sector overnight — it is that AI is allowing Western firms to re-onshore or redistribute white-collar work away from Indian outsourcing hubs entirely, quietly dismantling the labour-arbitrage model that underpinned three decades of middle-class expansion. India built its development story on a shortcut: skip mass manufacturing, sell English-speaking brainpower to Western back offices, and let the services sector do the work that factories did elsewhere. That shortcut is now the vulnerability.
The numbers from FY2023–24 make the inflection point visible. TCS, India’s largest private-sector employer in technology, reduced its workforce by 13,249 people to 601,546 even as revenues held up. Infosys, the sector’s bellwether for investor sentiment, saw headcount drop by 5.9% — its steepest annual decline on record. NASSCOM projects the industry could reach US$310–325 billion in revenue by FY2025–26, but the revenue trajectory and the employment trajectory are no longer moving together. That decoupling is the signal that matters.
The jobs most exposed — entry-level coding, software testing, routine customer support, and back-office processing — are precisely the roles that absorbed hundreds of thousands of fresh graduates each year and gave India’s demographic dividend its economic meaning. If those roles contract faster than new AI-adjacent roles emerge, the dividend becomes a liability.
RELATED
The details: where the pressure is landing
Infosys CEO Salil Parekh told analysts in April 2024 that generative AI was already deployed across “several hundred” client projects, improving productivity in application development, maintenance, and business operations — enabling clients to deliver the same output with the same or fewer people. That is a careful formulation, but its implication is clear: the unit of value is shifting from person-hours to outcomes.
Kris Gopalakrishnan, co-founder and former CEO of Infosys, was more direct. In a March 2024 interview he warned that routine coding and testing roles will shrink and that without rapid reskilling “we will have a large number of people who are not employable.” The warning carries weight precisely because it comes from someone with no incentive to alarm the market unnecessarily.
The specific job categories under pressure are not arbitrary. AI coding agents can now handle bug fixes, maintenance tasks, and multi-file engineering work — the entry-level functions that once absorbed armies of campus recruits. Call-centre operations face a parallel squeeze: AI chatbots already handle routine query resolution in India, and as implementation matures, the business case for large human support teams weakens structurally. The International Labour Organization has found that clerical and office tasks face higher AI exposure than most physical roles, which maps almost exactly onto India’s services export profile.
India’s Ministry of Electronics and Information Technology (MeitY) issued an advisory on 1 March 2024 requiring significant AI models to obtain government approval before public launch and to label outputs that are unreliable or under testing. The Digital Personal Data Protection Act, 2023, notified on 11 August 2023, establishes consent-based processing rules that directly affect AI systems handling personal data. Neither framework addresses AI’s labour market impact — a regulatory gap that is becoming harder to ignore as headcount data accumulates.
| Company | Headcount (March 2024) | Year-on-year change | Revenue trajectory |
|---|---|---|---|
| TCS | 601,546 | −13,249 (net reduction) | Continued growth |
| Infosys | 317,240 | −5.9% (largest annual decline in company history) | Continued growth |
How India arrived at this exposure
Most successful development economies followed a recognisable sequence: agriculture, then manufacturing, then higher-value services. India compressed that arc dramatically. Services now account for roughly 54% of GDP while employing around 31% of the labour force; agriculture contributes only 15% of GDP but still employs close to 46% of workers. The services-led model looked like a clever shortcut — English-speaking graduates, low labour costs, and Western multinationals desperate to cut back-office expenses. Instead of building a Shenzhen, India built the world’s software support desk.
That was a genuine accomplishment. It created a services export machine, a new urban middle class, and serious global credibility for Indian engineering. The problem is that the model’s comparative advantage — cheaper humans doing cognitive work — is precisely what generative AI competes with most directly. A factory worker in Pune is not immune to automation, but manufacturing still provides large numbers of people with a structured entry into regular employment. India never fully built that manufacturing base, and now AI is moving into white-collar services before the manufacturing alternative is in place.
India’s regulatory posture reflects the same tension. The country lacks a dedicated AI law; governance runs through the DPDP Act, the IT Act, and MeitY advisories focused on data protection and platform responsibility rather than risk classification or labour impact. The EU’s AI Act, published 12 July 2024, sets out detailed obligations by system risk level. India’s lighter approach reflects a deliberate choice to prioritise innovation and services exports — but it leaves accountability gaps that will become harder to manage as AI reshapes hiring at scale. You can review NASSCOM’s 2024 Strategic Review for the sector’s own baseline assessment of where the industry stands entering this transition.
Beyond the headline
The bigger picture
India’s AI shock is really a question of whether a development model built on exporting human brainpower can survive when cognition itself is partially automated. The country is discovering that skipping the low-tech manufacturing phase left it heavily exposed to a technology that competes directly with its core comparative advantage. This is not a cyclical hiring slowdown — it is a structural challenge to the sequencing of India’s entire growth model.
The reach
If AI erodes India’s dominance in outsourced services, Western firms face a more fragmented supplier base, higher transition costs, and new operational risks as work is redistributed or re-onshored. Investors holding India-focused IT funds, GDRs, or ADRs will need to distinguish sharply between firms that successfully monetise AI-enabled services and those whose legacy labour-arbitrage models get squeezed by hyperscaler automation sold directly to end clients.
Our take
AI is not secretly destroying India’s economy — it is openly destroying complacency about a guaranteed path from engineering degree to IT desk job. The firms and policymakers treating AI as a margin booster while deferring the social consequences are underestimating the adjustment ahead. India’s next decade will be defined less by whether it adopts AI and more by whether it uses the transition window to diversify beyond a single, services-heavy ladder to the middle class — because that window is not indefinitely open.
What this means for investors, employers, and policymakers watching India
With India’s two largest IT exporters already showing revenue-headcount decoupling in FY2023–24 data, and campus hiring under visible pressure, the next 12–18 months of quarterly results will determine whether this is a managed productivity shift or the beginning of a structural contraction in formal employment.
- Track quarterly utilisation rates, not just revenue: When TCS, Infosys, Wipro, and HCL report results through FY2025–26, watch employee utilisation rates and campus offer numbers alongside revenue. Rising utilisation with flat or falling headcount is the clearest signal that AI is replacing hiring rather than augmenting it.
- Reassess India-focused IT equity exposure: Investors in India-focused IT funds or individual ADRs should distinguish between firms building proprietary AI-managed service layers and those still primarily selling human hours. The valuation gap between these two cohorts is likely to widen through 2026–27.
- Monitor Karnataka’s Global Capability Centre targets: The state government has planned incentives to double global capability centres to 1,000 by 2029. GCC expansion offers higher-quality jobs but cannot absorb the broad base of graduates who previously entered conventional IT services — watching actual hiring against targets will indicate how much of the employment gap GCCs can realistically fill.
- Watch MeitY’s AI governance evolution: India’s current advisory-based approach to AI regulation leaves labour impact unaddressed. Any shift toward binding risk-based rules — prompted by EU AI Act alignment pressure or domestic political friction over graduate unemployment — would materially change compliance costs for both Indian IT firms and their Western clients.
- Consider supply-chain diversification risk: Western procurement teams relying heavily on Indian IT outsourcers should model scenarios in which pricing shifts from headcount-based to outcome-based contracts, and assess whether their vendor relationships are built for that transition or locked into legacy structures that become expensive to unwind.
FAQ
Which specific job categories in India’s IT sector face the highest risk from AI automation?
Entry-level software coding and testing, routine customer support, and back-office data processing face the most immediate pressure. These roles absorbed the largest share of campus recruits and are most exposed to AI coding agents and chatbot deployment. Mid-career roles in infrastructure management and software testing are also vulnerable. Senior engineering, AI deployment management, and client relationship functions are less immediately at risk.
How does India’s regulatory framework for AI compare to the EU AI Act?
India has no dedicated AI law. Governance runs through the Digital Personal Data Protection Act 2023, the IT Act, and MeitY advisories that focus on data consent and platform responsibility rather than risk classification. The EU AI Act, published 12 July 2024, sets binding obligations by system risk level. India’s lighter approach prioritises innovation and export competitiveness but leaves gaps on accountability, labour impact, and transparency that are likely to face political pressure as AI reshapes hiring.
What are Global Capability Centres, and can they replace traditional IT outsourcing jobs?
Global capability centres are in-house Indian operations built directly by multinationals — handling software development, analytics, finance, and R&D — rather than work contracted to firms like TCS or Infosys. GCC revenues are projected near US$100 billion with employment above 2 million people. They offer higher-quality roles but require specialist skills in data science, product management, and engineering. They cannot absorb the broad graduate base that conventional IT services employment once accommodated.
Is the headcount decline at TCS and Infosys directly caused by AI, or are other factors involved?
Both companies cite multiple factors: AI-driven productivity gains, weaker client demand from Western firms, skill mismatches, and tighter cost management. Separating genuine AI displacement from cyclical demand weakness or post-pandemic overhiring corrections is difficult. The significant signal is that revenues continued growing while headcount fell — a decoupling that suggests productivity gains, whether AI-driven or otherwise, are reducing the labour intensity of existing work rather than simply reflecting a revenue slowdown.
