AI acts as a general‑purpose productivity engine but with uneven effects: it can spur growth and innovation while amplifying job disruption and inequality, yet the evidence base is fragmented and context‑dependent, calling for more integrated and causal research.

Main Finding

Artificial intelligence (AI) acts as a general-purpose technology that can raise productivity, spur innovation, and support economic growth, but its developmental effects are highly uneven and context-dependent. The systematic review of 194 peer‑reviewed articles (2011–2025) finds positive potential across multiple development dimensions (productivity, trade, innovation), while also documenting trade‑offs including job displacement, widening inequality, environmental and energy challenges, and fragmented evidence due to measurement and scope limitations in the literature.

Key Points

• Scope and contribution
  • Systematic review of 194 Scopus‑indexed, peer‑reviewed articles (2011–2025) synthesizing AI’s impacts on labour markets, productivity, growth, inequality, trade, innovation, poverty, and environmental sustainability.
  • Emphasizes integration across development dimensions and situates AI within sustainable development frameworks (SDGs).
• Stylized findings by dimension
  • Labour markets: Evidence of both job displacement (particularly for routine tasks) and job creation/skill upgrading; redistributional pressures favoring higher‑skill workers.
  • Productivity & growth: AI linked to productivity gains and potential long‑run growth effects, but measurement challenges and heterogeneous firm/sectoral adoption lead to mixed empirical estimates.
  • Inequality & distribution: Diffusion of AI often associated with widening wage and regional disparities unless countervailed by policies and broad-based skill formation.
  • Innovation & competitiveness: AI accelerates innovation, knowledge accumulation, and firm competitiveness—especially where complementary capabilities (R&D, human capital, institutions) exist.
  • Trade & global value chains: AI reshapes comparative advantages, production organization, and GVC participation, with ambiguous effects on developing‑country integration.
  • Poverty & inclusion: Potential to reduce poverty through efficiency gains and new services, but benefits are uneven and may bypass marginalized groups.
  • Energy & environment: Mixed outcomes—AI can improve energy efficiency and environmental management, yet AI deployment and model training raise direct energy/carbon costs and lifecycle concerns.
• Research landscape and gaps
  • Rapid growth in publications, especially after 2020; earlier work (2011–15) was conceptual, 2016–19 saw more empirical work but indirect AI measures, and 2020 onward shows sharp expansion.
  • Literature is fragmented (theme‑by‑theme), geographically skewed toward advanced economies, and constrained by indirect AI measures, limited causal identification, and under‑studied environmental and distributional consequences.

Data & Methods

• Search and selection
  • Database: Scopus (peer‑reviewed journal articles only).
  • Timeframe: 2011–2025 (captures modern ML/deep learning era).
  • Initial hits: 1,194 records → screening (title/abstract/full text) → 194 final articles included.
  • Search strategy: (“Artificial Intelligence” OR “AI”) AND development‑related terms (economic development, productivity, labor market, trade, innovation, inequality, energy, sustainability).
  • Language: English only.
• Screening & coding
  • Screening managed with Covidence and summarized via PRISMA flow.
  • Coding framework captured bibliographic info, region, methodology (empirical, review, modelling), primary theme (labour, productivity, trade, inequality, innovation, poverty, environment), theoretical lens, and key findings.
  • Thematic classification used an inductive–deductive approach; each article assigned to the dominant theme while noting multidimensional content.
• Limitations of the review method
  • Restriction to Scopus and peer‑reviewed articles excludes working papers, reports (World Bank, OECD), preprints (SSRN, arXiv), and non‑English studies.
  • AI measurement in the underlying literature is often indirect (robots, patents, digital proxies), and many primary studies rely on associational rather than causal designs—issues reflected in the review.

Implications for AI Economics

• For research
  • Improve measurement: develop and standardize direct exposure metrics for AI (task‑level, model usage, firm‑level deployment) and combine with granular labour and output data.
  • Causal inference: prioritize quasi‑experimental and longitudinal designs (natural experiments, difference‑in‑differences, instrumenting adoption) to identify causal effects on employment, productivity, and distribution.
  • Integrative, multisectoral frameworks: bridge fragmented literatures by studying joint pathways (e.g., how productivity gains interact with labour markets, trade, and environmental outcomes).
  • Focus on heterogeneity: examine sectoral, firm‑size, regional, and country‑income heterogeneity—especially for low‑ and middle‑income countries where institutional complementarities matter.
  • Environmental lifecycle analysis: quantify AI’s direct and indirect energy/carbon footprints and trade‑offs between digitalization gains and sustainability.
  • Broaden evidence base: incorporate working papers, policy reports, and non‑English studies to reduce geographical and publication bias.
• For policy and development practice
  • Active labour and education policies: invest in reskilling/upskilling, STEM and digital literacy, and lifelong learning to mitigate displacement and enable complementary skills.
  • Inclusive diffusion: support small and medium enterprises and lagging regions to adopt AI via finance, extension services, and technology transfer to avoid widening gaps.
  • Social protection and redistribution: strengthen safety nets, wage policies, and progressive taxation to address distributional impacts.
  • Governance and regulation: develop standards for data governance, algorithmic accountability, and competition policy that consider development objectives.
  • Sustainability governance: integrate AI policy with energy and environmental strategies to ensure net environmental benefits (e.g., incentivize energy‑efficient models, green data centers).
  • International cooperation: assist developing countries in building institutional capacity, data infrastructure, and regulatory frameworks so AI’s benefits are more broadly shared.

Reference: Ho, S.-Y., & Giwa, F. (2026). Artificial Intelligence and Economic Development: A Systematic Review of Patterns and Pathways. Journal of the Knowledge Economy. https://doi.org/10.1007/s13132-026-03385-w