AI is not an automatic growth engine: its economic payoff depends on human capital, data infrastructure and institutional complements, causing similar AI investments to yield divergent outcomes across countries and firms.

Main Finding

AI—treated as a general-purpose technology—does not automatically generate uniform or large macroeconomic growth effects. The realized contribution of AI to growth and structural change is conditional on an ecosystem of complementary factors (human capital, digital/data infrastructure, institutional coordination, governance). Interpreting task-based automation alongside endogenous-growth and open-innovation frameworks yields an ecosystem-centered, conditional account: similar AI investments can produce divergent structural outcomes depending on institutional and structural context.

Key Points

• Conceptual framing: AI is analyzed as a general-purpose technology (GPT) whose economy-wide effects unfold through multiple, interacting channels rather than a single deterministic pathway.
• Three core transmission channels identified:
• Total factor productivity (TFP) — productivity gains from automation, process improvements, and AI-enabled complementarities.
• Task reallocation and labor-market restructuring — substitution and augmentation across tasks, changing occupational and sectoral composition.
• Innovation and knowledge-generation dynamics — AI’s influence on R&D productivity, idea flows, open innovation, and cumulative knowledge creation.
• Methodological stance: a structured, mechanism-oriented literature synthesis (2015–2025) using transparent search, screening, and thematic classification; not a formal meta-analysis.
• Empirical evidence is heterogeneous: estimates of AI’s macroeconomic contribution vary substantially across studies, countries, sectors, and institutional settings.
• Key mechanism: institutional and structural complementarity. Outcomes depend more on enabling conditions (skills, data/digital infrastructure, governance, coordination) than on AI technology alone.
• Theoretical integration: Combining task-based automation models with endogenous growth and open-innovation perspectives explains why identical AI inputs can yield different growth and structural-change trajectories.
• Contribution to literature: (i) distinguishes foundational theoretical perspectives from recent empirical evidence, (ii) clarifies the link between task transformation and broader structural transformation, and (iii) highlights institutional complementarity as central to AI-driven development.

Data & Methods

• Scope: systematic literature review of studies published 2015–2025.
• Approach: structured, mechanism-oriented synthesis emphasizing transparent search, screening, and thematic classification protocols; aimed at identifying transmission mechanisms and contextual moderators.
• Not a meta-analysis: the study does not pool effect sizes or follow formal meta-analytic statistical protocols; it synthesizes mechanisms and patterns qualitatively and conceptually.
• Outcome: development of an integrative innovation-ecosystem framework that links the three transmission channels (TFP, task reallocation, innovation dynamics) and foregrounds complementary conditions and institutional mediators.

Implications for AI Economics

• Modeling implications:
  • Move beyond one-dimensional causal assumptions; incorporate conditionality and interaction terms for institutional complements (skills, infrastructure, governance) in macro and micro models.
  • Integrate task-based labor models with endogenous-growth and open-innovation mechanisms to capture feedback between productivity, knowledge creation, and structural change.
• Measurement and empirical strategy:
  • Prioritize cross-country and within-country heterogeneity analysis; collect data on data-capital, digital infrastructure, institutional quality, and skill composition.
  • Use longitudinal microdata, matched employer–employee panels, and firm-level R&D/innovation measures to trace mechanistic pathways.
  • Design quasi-experimental and structural-counterfactual studies to separate substitution vs. augmentation and short-run vs. long-run effects.
• Policy implications:
  • Policies should focus on building complements (education and reskilling, data governance, digital infrastructure, coordinated innovation policy) rather than treating AI investment alone as sufficient.
  • Institutional coordination and governance matter for enabling positive diffusion and limiting adverse distributional outcomes.
  • Active labor-market policies and social safety nets are needed to manage transitional dislocations from task reallocation.
• Research gaps:
  • Better causal identification of AI’s macroeffects across diverse institutional contexts.
  • Quantification of complementarities between AI and non-AI inputs (human capital, data capital, institutions).
  • Understanding long-run feedbacks between AI-enabled innovation, market structure, and knowledge spillovers.
• Practical takeaway: AI’s promise for growth is real but conditional—economists and policymakers should evaluate AI as part of an innovation ecosystem and design interventions that strengthen the necessary complements for inclusive and sustained structural transformation.