AI adoption raises Tunisian SMEs' financial and operational performance, but the returns concentrate in firms with stronger finance and technical skills; weak formal institutions do not significantly blunt these gains, implying firms use adaptive informal mechanisms.
Structural Constraints as Moderators in the Ai–performance Relationship: Evidence from Smes in an Emerging Economy
In a survey of 280 Tunisian SMEs, self-reported AI adoption is associated with improved financial and operational performance, with the performance gains amplified by firms' financial capacity and technical competencies but not significantly moderated by institutional quality.
Artificial intelligence (AI) adoption is increasingly recognized as a source of competitiveness for small and medium-sized enterprises (SMEs). Yet, prior research has primarily treated structural constraints such as financial scarcity, skill shortages, and institutional weaknesses as mere barriers, leaving their post-adoption impact underexplored, particularly in emerging economy contexts. This study empirically examines a relationship that has been theoretically acknowledged but rarely tested in such settings. Drawing on the Resource-Based View, Contingency Theory, and Institutional Theory, we propose a multidimensional framework explaining why AI adoption does not uniformly translate into performance gains but depends on firms’ financial capacity, technical competencies, and institutional environment. Using survey data from 280 Tunisian SMEs analyzed with partial least squares structural equation modeling (PLS-SEM), results confirm that AI adoption significantly improves financial and operational performance. Financial and technical strengths amplify these effects, whereas institutional conditions exert no significant moderating influence, suggesting that firms compensate for institutional weaknesses through adaptive and informal mechanisms. By reconceptualizing structural constraints as post-adoption moderators rather than pre-adoption barriers, this study advances understanding of contextual contingencies shaping AI outcomes and provides insights for managers and policymakers in resource-limited economies.
Summary
Main Finding
AI adoption (measured ex-post as embedded use) significantly improves both financial and operational performance of Tunisian SMEs. The positive effects are amplified when firms have stronger financial capacity and technical competencies; institutional conditions showed no significant moderating effect.
Key Points
- The paper reframes structural constraints (financial scarcity, technical skill gaps, institutional weakness) as post-adoption moderators rather than solely pre-adoption barriers.
- Theoretical grounding: Resource-Based View (RBV), Contingency Theory (CT), and Institutional Theory (IT) — AI is a strategic resource whose value realization depends on complementary assets, fit with organizational capabilities, and the institutional environment.
- Hypotheses:
- H1: AI adoption positively affects SME financial and operational performance — supported.
- H2–H4: Financial, technical, and institutional constraints negatively moderate the AI → performance link — results: financial and technical constraints (i.e., lack of strength) materially weaken AI gains (so financial/technical strength amplifies gains); institutional constraints did not significantly moderate the relationship.
- Interpreted result on institutions: firms appear to deploy adaptive/informal mechanisms to compensate for weak institutional environments, which may explain the non-significant institutional moderation.
Data & Methods
- Context: Tunisia (an emerging economy), focus on SMEs across manufacturing, services, trade/retail, and ICT.
- Sample: 280 valid responses collected June–August 2025 via a purposive, non-probability survey of senior executives/technical/financial managers in four hubs (Greater Tunis, Sousse, Nabeul, Sfax). Composition: ~59% small (10–49 employees), 41% medium (50–249); sector shares reported.
- Measurement: AI adoption operationalized as ex-post embedded use (readiness → implementation → institutionalization); structural constraints operationalized as multidimensional constructs: financial access/capacity, technical infrastructure/skills, and institutional support/stability. Performance measured on financial and operational dimensions (self-reported).
- Analysis: Partial Least Squares Structural Equation Modeling (PLS-SEM) to test direct and moderating effects.
- Methodological positioning: positivist, hypothetico-deductive. (Limitations implied by cross-sectional, self-reported data and purposive sampling.)
Implications for AI Economics
- Theory
- Advances theory by shifting attention from adoption decisions to post-adoption outcomes: structural constraints determine whether AI becomes a strategic, value-generating capability.
- Reinforces RBV and contingency perspectives: complementary assets (finance, technical skills) are necessary to extract value from AI; fit matters.
- Suggests institutional theory’s effects may be more complex in resource-limited settings where firms use informal/adaptive strategies, calling for more nuanced operationalization of institutional influence in AI studies.
- Empirical research
- Encourages more ex-post, capability-oriented measurements of AI adoption (institutionalization, routinization) rather than binary adoption indicators.
- Calls for longitudinal and probability-based studies to track dynamic complementarities (investment → learning → performance) and to validate the non-significant institutional finding across contexts.
- Policy and practice
- For policymakers in emerging economies: prioritize interventions that expand SME access to finance (grants, concessional loans, blended finance) and build technical capacity (training, subsidized cloud/analytics infrastructure). These will have larger marginal returns on AI-enabled performance than focusing solely on top-down regulation.
- For managers: invest in complementary assets (data systems, staff upskilling, maintenance/upgrade budgets) to amplify AI returns; consider adaptive/informal institutional strategies (partnerships, local networks) where formal support is weak.
- Broader economic implications
- Differential diffusion of AI across firms within emerging economies may widen productivity gaps unless complementary finance and technical capacity constraints are addressed.
- Policies that lower the cost of maintaining and scaling AI (finance and skills) can improve the aggregate payoff to AI investments and support inclusive digital transformation.
Limitations to note for interpretation: cross-sectional self-reports, purposive sampling in one country, and possible heterogeneity across sectors not fully generalizable; further longitudinal and multi-country research is recommended.
Assessment
Paper Typecorrelational
Evidence Strengthlow — Findings rely on cross-sectional, self-reported survey data from a modest sample (N=280), creating risk of reverse causality, omitted-variable bias, and common-method variance; no quasi-experimental or instrumental strategy is used to support causal claims.
Methods Rigormedium — PLS-SEM is an appropriate exploratory/confirmatory technique for modeling latent constructs and interaction paths with relatively small samples and theory-driven hypotheses; the authors note robustness checks, but the lack of longitudinal data, external validation of measures, and exogenous identification limits methodological rigor for causal inference.
SampleCross-sectional survey of 280 small and medium-sized enterprises (SMEs) in Tunisia; key measured variables are self-reported AI adoption, financial performance, operational performance, financial capacity, technical competence, and perceptions of the institutional environment; sampling frame and sectoral breakdown not specified in the summary.
Themesproductivity adoption skills_training governance
IdentificationCross-sectional survey of SMEs with partial least squares structural equation modeling (PLS-SEM) to estimate associations and moderation effects between self-reported AI adoption and firm outcomes; no exogenous variation, instruments, panel data, or randomized assignment—identification is associative rather than causal.
GeneralizabilitySingle-country study (Tunisia) — findings may not generalize to other institutional or economic contexts, SME-only sample — results may not apply to large firms or multinational enterprises, Cross-sectional, self-reported data — limits external validity and causal claims, Modest sample size and unspecified sampling frame — potential non-representativeness across sectors and firm types, Context-specific coping mechanisms (informal networks) may differ across cultures and regulatory regimes
Claims (9)
| Claim | Direction | Confidence | Outcome | Details |
|---|---|---|---|---|
| AI adoption significantly improves financial performance of Tunisian SMEs. Firm Revenue | positive | medium | financial performance |
n=280
0.09
|
| AI adoption significantly improves operational performance of Tunisian SMEs. Firm Productivity | positive | medium | operational performance |
n=280
0.09
|
| Firms' financial capacity amplifies the positive effect of AI adoption on performance. Firm Productivity | positive | medium | AI adoption → (financial and/or operational) performance (moderated by financial capacity) |
n=280
0.09
|
| Firms' technical competencies amplify the positive effect of AI adoption on performance. Firm Productivity | positive | medium | AI adoption → (financial and/or operational) performance (moderated by technical competencies) |
n=280
0.09
|
| Institutional conditions do not exert a significant moderating influence on the relationship between AI adoption and firm performance in this sample. Firm Productivity | null_result | medium | AI adoption → performance (moderated by institutional conditions) |
n=280
0.09
|
| Firms compensate for institutional weaknesses through adaptive and informal mechanisms, allowing AI adoption to yield performance gains despite weak institutions. Firm Productivity | positive | low | firm-level compensatory/adaptive mechanisms enabling AI-related performance gains (inferred, not directly measured) |
n=280
0.04
|
| Reconceptualizing structural constraints as post-adoption moderators rather than pre-adoption barriers improves understanding of contextual contingencies shaping AI outcomes in resource-limited economies. Research Productivity | mixed | low | theoretical understanding of how structural constraints operate (conceptual/outcome: explanatory framework validity) |
n=280
0.04
|
| The study provides actionable insights for managers and policymakers in resource-limited economies regarding factors that influence whether AI adoption translates into performance gains. Firm Productivity | positive | low | practical guidance/implications for managerial and policy decision-making (inferred from empirical findings) |
n=280
0.04
|
| This study empirically tests a theoretically acknowledged but rarely tested relationship (AI adoption → performance conditional on structural constraints) in an emerging-economy setting. Firm Productivity | null_result | medium | existence and nature of the conditional relationship between AI adoption and firm performance in an emerging economy (empirical test performed) |
n=280
0.09
|