AI raises productivity at the firm level by automating tasks and augmenting decision-making, but those gains rarely translate into uniform macroeconomic improvement because slow diffusion, skill gaps and weak data infrastructure limit scaling; without reskilling and organizational change, AI risks widening productivity and regional divides.

Main Finding

AI functions as a powerful productivity-enhancing general-purpose technology at the firm and task level—delivering substantial labor productivity gains through both automation and augmentation—but macroeconomic productivity effects are uneven and limited so far. Realized gains depend critically on complementary investments (skills, data infrastructure, organizational redesign), national preparedness, and communicative/strategic implementation; without these, AI can widen productivity, skill, and regional gaps.

Key Points

• Theoretical frames
  • Neoclassical/endogenous growth: AI as productivity-enhancing capital that can raise output per worker, but long-run gains require complementary investments and knowledge spillovers.
  • Task-based frameworks: Impact depends on task content—routine tasks are automatable, non-routine cognitive/interpersonal tasks are more likely to be augmented.
  • SBTC vs RBTC: AI contributes to skill polarization; routine-biased automation especially threatens middle-skill, routine cognitive jobs.
  • GPT perspective: AI has GPT characteristics (like electricity/ICT) but initial diffusion yields asymmetric benefits between leaders and laggards.
• Organizational/management insights
  • Human–AI collaboration models often increase productivity by augmenting decision-making; mixed-intelligence teams combine human contextual judgment with AI speed/scale.
  • Full productivity gains typically require workflow re-engineering, digital transformation, and managerial change, not just tool deployment.
  • Communication (internal and external) shapes acceptance, reskilling uptake, and implementation speed.
• Empirical evidence (emerging, heterogeneous)
  • Industry: Manufacturing sees visible gains via robotics/vision/automation; service sectors (finance, healthcare, logistics, customer service) show productivity and quality gains, often by freeing humans for more complex work.
  • Firm-level: Adopters, especially large firms with resources for infrastructure and training, show higher labor productivity; SMEs face barriers (costs, data, skills) that slow gains and widen productivity gaps.
  • Worker-level: Tools (including generative AI) can raise individual task speed/accuracy—even for lower-skilled workers—though risks include overreliance and potential long-term skill erosion.
  • Regional/macro: National AI preparedness (infrastructure, human capital, regulation) strongly conditions outcomes; regulatory uncertainty and digital divides hinder diffusion.
• Robust gaps identified
  • Much evidence is cross-sectional/short-term; longitudinal, causal, sectoral, and informal-economy studies are scarce.
  • Understudied sectors: education, public administration, small retail, and informal labor markets in developing countries.
  • Need for integrated research linking micro adoption processes to macro productivity dynamics.

Data & Methods

• Approach: Narrative literature review synthesizing interdisciplinary scholarship (economics, management, labor studies, business/economic communication) published between 2010 and 2025.
• Sources and evidence types summarized: theoretical models, task-based analyses, firm- and industry-level empirical studies (case studies, panel/cross-sectional analyses where available), worker-level experiments/field studies on AI tools (including generative AI), and country/regional preparedness assessments.
• Methodological limitations noted by the authors: reliance on mostly short-term and cross-sectional empirical work, heterogeneity in measurement of AI adoption and productivity, limited causal identification in many studies, and uneven geographic/sectoral coverage.

Implications for AI Economics

• For researchers
  • Prioritize longitudinal and causal studies that trace AI diffusion and complementary investments over time.
  • Expand cross-country, sectoral, and informal-economy research to understand heterogeneity in diffusion and impacts.
  • Investigate worker skill dynamics (reskilling, deskilling risks) and firm-level complementarities (organization, data assets).
• For managers and firms
  • Treat AI adoption as a socio-technical transformation: invest in data infrastructure, reskilling/upskilling, workflow redesign, and change management to capture productivity gains.
  • SMEs need targeted support (affordable platforms, training, data-sharing arrangements) to avoid widening productivity gaps.
  • Emphasize transparent communication to build trust, encourage reskilling, and reduce resistance.
• For policymakers
  • Support complementary public investments (digital infrastructure, education and retraining programs, R&D) to enable broad-based productivity benefits.
  • Clarify regulation on data, liability, and ethics to reduce adoption uncertainty.
  • Design inclusive policies that mitigate distributional consequences (skill polarization, regional divergence), e.g., targeted subsidies, SME assistance, regional capacity building.
• Broader economic considerations
  • Short-run firm-level gains do not automatically translate to macro productivity growth; diffusion, absorptive capacity, and institutional complements are decisive.
  • Distributional effects (early-adopter advantage, wage inequality, skill polarization) should be part of macroeconomic assessments of AI’s role in growth.

Reference: Olatunbosun et al. (2025), "How Artificial Intelligence Is Reshaping Labor Productivity: A Narrative Review…" Journal of Economics, Business, and Commerce, 2(2), 414–425.