AI adoption correlates with sizable gains in manufacturing output and greener operations among SMEs in Yemen and Saudi Arabia, but benefits hinge on infrastructure and institutional support; Saudi firms benefit from advanced technologies and institutions while Yemeni firms rely on low-cost adoption and organizational flexibility.

Main Finding

Integrating AI into a Technology–Organization–Environment (TOE) framework (an “AI-TOE”) significantly improves manufacturing performance (β = 0.759) and environmental performance (β = 0.487) for SMEs in fragile and rapidly transforming economies. Manufacturing (industrial) performance acts as an important mediator: AI-enabled TOE boosts operational efficiency, which in turn accelerates environmental gains. The effects differ by context — Saudi SMEs benefit more from institutional support and advanced infrastructure, while Yemeni SMEs rely on low‑cost AI solutions and organizational flexibility (“survival-oriented innovation”).

Key Points

• Core contribution: an AI-enhanced TOE framework that (a) embeds AI capabilities across the three TOE dimensions, (b) introduces “survival-oriented innovation” for fragile contexts, and (c) describes “institutional leapfrogging” as a pathway for rapid transformers.
• Empirical results (PLS-SEM):
  • Direct positive effect of AI-TOE on manufacturing performance: β = 0.759.
  • Direct positive effect of AI-TOE on environmental performance: β = 0.487.
  • Manufacturing performance mediates the AI → environmental performance link.
• Contextual differences:
  • Saudi Arabia: advanced infrastructure and institutional support enable higher-value AI applications (predictive maintenance, cloud ERP, analytics) and faster institutional adaptation (linked to Vision 2030).
  • Yemen: extreme fragility leads firms to adopt lower-cost, simpler AI (chatbots, inventory forecasting); organizational flexibility and survival strategies drive impact despite infrastructure gaps.
• Conceptual/academic advances: first holistic AI-integrated TOE tailored for fragile/transforming economies; extends TOE theory for resource-constrained settings and links AI adoption to both operational and environmental performance simultaneously.
• Policy/managerial takeaways: prioritize context-appropriate AI investments (low-cost, high-impact tools in fragile states; deeper integration where infrastructure and institutions permit); pair AI adoption with institutional support, training, and environmental regulation alignment.

Data & Methods

• Sample and design:
  • Frame: 1,006 SMEs identified across Yemen and Saudi Arabia; two-wave field survey.
  • Final usable responses: 294 (response rate 29.2%), balanced by country (Yemen n=147; Saudi Arabia n=147).
  • Sampling covered urban and rural firms across several governorates/cities in both countries.
• Measurement and instrument:
  • Questionnaire adapted from prior literature, translated/back-translated to Arabic, reviewed by regional AI researchers, piloted (excluded pilot responses).
  • Respondent selection via CEOs who recommended knowledgeable employees; two data waves to reduce common-method bias.
• Analysis:
  • Partial Least Squares Structural Equation Modeling (PLS-SEM) using SmartPLS v3.2.7; SPSS for auxiliary tests.
  • Measurement validity: Composite Reliability (CR) > 0.70; Cronbach’s α acceptable (>0.50 threshold used); Average Variance Extracted (AVE) > 0.50; cross-loadings and Fornell–Larcker checks satisfied discriminant validity.
  • Common-method and multicollinearity checks: Harman’s first factor = 35% (below 50% concern threshold); VIF range 1.170–2.595 (no multicollinearity).
• Hypotheses (tested): AI-supported TOE positively influences environmental performance (H1); manufacturing performance mediates and positively influences environmental performance (H2–H4); AI-TOE directly enhances manufacturing performance.

Implications for AI Economics

• Theory and modeling:
  • Demonstrates how integrating AI into established adoption frameworks (TOE) changes performance dynamics in low-resource/fragile settings — relevant for models of technology diffusion and productivity in heterogeneous economies.
  • Suggests path-dependence and institution-infrastructure interactions: the same AI capabilities produce different economic and environmental returns depending on institutional maturity and infrastructure — important for micro-founded models of technology adoption and productivity spillovers.
• Policy and developmental economics:
  • Supports targeted public investments: strengthening institutional support and basic digital infrastructure amplifies the welfare and environmental benefits of AI for SMEs.
  • Cost-effectiveness in fragile contexts: low-cost, modular AI interventions can yield survival and efficiency gains; development policy should consider staged AI adoption strategies rather than “big-bang” investments.
• Measurement and empirical research:
  • Validates use of PLS-SEM and composite reliability approaches for multi-dimensional, context-sensitive adoption studies in emerging markets; highlights need for longitudinal work to capture dynamic institutional leapfrogging.
  • Raises the prospect of an “AI Environmental Kuznets Curve” in rapidly transforming economies — opening empirical work on nonlinear relationships between AI adoption, productivity growth, and environmental externalities.
• Recommendations for future research in AI economics:
  • Conduct longitudinal/panel studies to unpack causal dynamics, technology maturation, and potential rebound effects.
  • Disaggregate AI types and applications (e.g., predictive maintenance vs. generative AI) to estimate heterogeneous returns to firms and environmental impacts.
  • Extend cross-country comparisons across more fragile and transition economies to test generalizability and identify policy levers that maximize inclusive gains.

Limitations to note (from the study): cross-sectional survey design with a single-respondent approach (mitigations applied but residual bias possible); moderate response rate (29.2%); country-pair comparison is illustrative but not exhaustive for all fragile/transforming economies.