Uneven AI adoption risks widening capability inequality: low-income groups and segmented labour markets—notably in India’s IT services—face accelerated skill-change and transition risks, calling for coordinated policies on learning equity, transition protections and accountable algorithmic management.

Main Finding

AI amplifies existing capability inequalities—unequal access to education, digital infrastructure, compute, and organizational adoption—so that productivity and growth gains concentrate where complements are strong; this produces distributional risks (task transformation, skill volatility, and wage polarization) rather than uniform job loss, with acute implications for segmented labor markets such as India’s and for the IT/services workforce.

Key Points

• Conceptual frame: “capability inequality” (skills + infrastructure + compute + adoption + governance) is the primary channel through which AI affects distributional outcomes.
• Global diffusion: Firm-level AI adoption rose sharply in recent years (OECD: 8.7% in 2023 → 14.2% in 2024 → 20.2% in 2025), but adoption is uneven across firms/countries.
• Divergence risk: UNDP and WTO warn that uneven readiness can produce a “Next Great Divergence” where low-capability countries or groups face disruption without commensurate gains.
• Labor-market transformation: ILO’s exposure measures and WEF forecasts indicate large task transformation (not uniform destruction), with employers expecting ~39% of key job skills to change by 2030—implying major reskilling needs and skill-volatility risks.
• India context: Macroeconomic growth remains robust (NSO/MoSPI real GDP estimates: FY24 9.2% revised → FY25 6.5%), PLFS shows high LFPR (60.1%) and WPR (58.2%), and HCES indicates a recent decline in consumption Gini —but labor-market segmentation and digital capability gaps make income pathways vulnerable.
• IT/services sector: Reuters reporting and firm signals (e.g., TCS) indicate AI may compress billable hours and alter billing models, creating internal restructuring—Nasscom still reports net hiring, implying reallocation and higher skill thresholds rather than pure net job loss.
• Policy direction: Prioritize learning equity (modular reskilling with measurable wage outcomes), transition protections (portable benefits, retraining-linked income support), accountable algorithmic management (explainability, appeals, audits), protections for entry-level ladders (apprenticeships, industry-aligned skilling), and distribution-sensitive metrics beyond GDP.

Data & Methods

• Design: Integrative secondary-data synthesis using a socio-technical political-economy lens. The study triangulates institutional reports, official Indian statistics, and sector evidence to map mechanisms and plausible distribution pathways; it does not claim causal identification.
• Principal data sources:
  • Global institutions: OECD (AI adoption, wage analysis), ILO (generative-AI occupational exposure index), UNDP (divergence analysis), WTO (World Trade Report 2025), WEF (Future of Jobs 2025).
  • India official: NSO/MoSPI GDP estimates; PLFS 2023–24 (LFPR, WPR, UR); HCES 2022–23 & 2023–24 (consumption Gini).
  • Sector evidence: Reuters reporting on IT services (2026), Nasscom Strategic Review (employment projections), India AI Impact Summit 2026 outputs, NITI Aayog roadmaps.
• Analytical procedure:
  • Establish macro and distributional baseline for India (GDP, PLFS, HCES).
  • Map capability channels (skills, connectivity, compute, adoption, governance).
  • Use ILO/WEF to frame occupational exposure and expected skill change.
  • Perform an IT-sector stress test using Reuters/Nasscom signals to infer internal role-restructuring risks.
  • Synthesize evidence into propositions and policy recommendations.
• Limitations:
  • Secondary-source synthesis without primary data collection or causal identification.
  • Rapidly evolving AI landscape means later firm-level or microdata could refine exposure estimates.
  • Sector signals (e.g., layoffs, billing model change) are indicative but not definitive causal attributions to AI.

Implications for AI Economics

• Distributional modelling: Economic models of AI should incorporate heterogeneity in complementary capabilities (education, broadband, compute access, firm adoption) rather than treating technology as neutral. This implies multi-layered general-equilibrium frameworks with endogenous skill formation and firm adoption margins.
• From automation to transformation: Research and policy should shift emphasis from aggregate job counts to task-level transformation, skill-bundling, and within-occupation inequality—model labor markets with changing task mixes, reallocation frictions, and heterogeneous reskilling costs.
• Measuring impacts: Monitoring must move beyond GDP to distribution-sensitive indicators (earnings stability, hours volatility, skill mobility, subgroup AI access/adoption, sectoral wage polarization). Integrate household surveys (HCES, PLFS) with administrative and firm-level AI-adoption data for more granular inference.
• Policy design: Targeted interventions (modular reskilling tied to labour-market outcomes, income support during transitions, apprenticeship incentives) can mitigate inequality amplification. Algorithmic governance (explainability, appeals, audits) is both a redistribution and a market-design tool—affecting bargaining power and matching.
• Sectoral focus — IT/services: The Indian IT model illustrates how AI can compress routine billable work while expanding demand for higher-end AI/data/cloud/security skills. Policy and firms must manage entry-level pipeline risks (apprenticeships, funded internships, industry-academia pathways) to preserve mobility channels for households.
• International cooperation: To reduce divergence, global public goods (digital infrastructure finance, shared compute access, capacity building, interoperable governance norms) are economically important; development models and trade policies should account for uneven AI readiness.
• Research priorities: Empirical microstudies linking firm-level AI adoption to wage trajectories, task-content change, hiring practices, and career ladders; experiments on reskilling program design and on algorithmic-management regulation; cross-country work on whether compute- and data-access interventions reduce divergence.
• Equity-by-design: Economists and policymakers should treat AI as a productivity regime requiring coordinated investments in human capital, infrastructure, and institutions to ensure efficiency gains translate into broad-based welfare improvements rather than concentrated rents.