MSMEs with stronger AI capabilities report better triple-bottom-line outcomes, largely because AI strengthens sensing, seizing and reconfiguration abilities that enable sustainable business-model innovation; direct effects of AI exist but are smaller than the mediated pathway.

Main Finding

AI capability functions as a higher‑order strategic/dynamic capability in MSMEs: it strengthens sensing–seizing–reconfiguring dynamic capabilities, which enable Sustainable Business Model Innovation (SBMI) and thereby improve Triple Bottom Line (TBL) outcomes (economic, environmental, social). AI has smaller but significant direct effects on TBL dimensions; however the largest impacts occur via sequential mediation through dynamic capabilities and SBMI.

Key Points

• Theoretical framing: integrates Resource‑Based View (RBV), Dynamic Capability Theory (DCT), and the Triple Bottom Line (TBL) framework; treats AI capability as a strategic resource that enables dynamic capabilities.
• Main hypothesized paths tested and results (standardized β, t-value):
  • AI Capability → Dynamic Capabilities: β = 0.54, t = 8.72 (supported)
  • Dynamic Capabilities → SBMI: β = 0.48, t = 7.15 (supported)
  • SBMI → Economic Performance: β = 0.42, t = 6.38
  • SBMI → Environmental Performance: β = 0.51, t = 7.91 (largest SBMI → TBL effect)
  • SBMI → Social Performance: β = 0.39, t = 5.84
  • Direct AI → Economic: β = 0.21, t = 3.66; AI → Environmental: β = 0.18, t = 2.94; AI → Social: β = 0.17, t = 2.61 (all supported but smaller)
• Mediation: significant indirect and sequential mediation:
  • AI → DC → SBMI: indirect β = 0.26 (p < 0.001)
  • DC → SBMI → Environmental Performance: indirect β = 0.24 (p < 0.001)
  • AI → DC → SBMI → TBL dimensions: significant sequential mediation (p < 0.001)
• Reliability and validity: Cronbach’s α, Composite Reliability > 0.85 for all constructs; AVE > 0.59; factor loadings > 0.65; Fornell–Larcker and HTMT criteria satisfied.
• Model fit: χ²/df ≈ 2.1–2.24, CFI ~0.92, TLI ~0.91–0.92, RMSEA ~0.056–0.061, SRMR ~0.049–0.052 — acceptable fit.
• Common method bias: Harman single‑factor = 28% (argued not a major concern); VIFs between 1.42–2.31.

Note on data reporting: the manuscript contains inconsistent statements about context/sample in places (mentions South Africa and India; sample sizes reported as 216 and 348). The methods and final analysis indicate 348 valid surveys of Indian MSMEs across manufacturing and services; treat 348 as the analytic sample.

Data & Methods

• Design: cross‑sectional survey; quantitative SEM analysis using SmartPLS 3.0.
• Sample: 348 valid firm‑level responses from MSMEs (paper reports distribution across Indian states: Karnataka, Maharashtra, Tamil Nadu, Gujarat, Delhi NCR); respondents held strategic roles (owner/operator, MD, head operations/technology, sustainability manager).
• Survey: ~30 items, 5‑point Likert scale; multi‑item validated scales for:
  • Artificial Intelligence Capability (data infrastructure, analytics, automation, governance)
  • Dynamic Capabilities (sensing, seizing, reconfiguring)
  • Sustainable Business Model Innovation (value proposition redesign, circular processes, stakeholder integration, resource efficiency)
  • TBL outcomes (economic, environmental, social subscales)
• Analysis:
  • Measurement model: CFA, Cronbach’s α, CR, AVE, Fornell‑Larcker, HTMT.
  • Structural model: path coefficients, t‑values, model fit indices.
  • Mediation: bootstrapping with 5,000 resamples to estimate indirect/sequential mediation.
• Limitations acknowledged by authors: cross‑sectional design (limits causal inference), geographic focus on Indian MSMEs, and suggestion for longitudinal and cross‑country follow‑ups.

Implications for AI Economics

• Mechanism matters: Economic returns to AI in MSMEs are delivered largely through capability development and business‑model transformation, not merely technology adoption. Economic analyses should account for the costs and benefits of complementary investments (training, processes, governance, data infrastructure), not only AI software/hardware costs.
• Policy and subsidy design: Public programs aiming to boost MSME productivity with AI should fund capability building (skills, organizational routines, managerial processes) and support SBMI (e.g., circular economy pilots), because these magnify AI’s TBL impacts—especially environmental outcomes.
• Prioritization for investors and managers: Expect stronger environmental payoff from AI-enabled SBMI (largest coefficient), so ESG‑oriented investors may obtain outsized non‑financial returns when supporting AI + SBMI strategies. Direct AI interventions (predictive maintenance, energy optimization) yield measurable direct economic/environmental gains but are smaller than mediated effects.
• Measurement and evaluation: Cost‑benefit and ROI studies of AI in small firms should incorporate intermediate capability and business‑model change variables to avoid underestimating long‑run value creation. Empirical models in AI economics should include dynamic capability constructs to better predict TBL outcomes.
• Research agenda for AI economics:
  • Longitudinal causal estimates of AI → dynamic capabilities → performance (panel data, quasi‑experimental designs).
  • Cross‑country and sectoral heterogeneity in mediated returns (are mediated effects larger in manufacturing vs. services, or across institutional environments?).
  • Quantify investment thresholds: how much capability‑building investment is required to unlock mediated TBL gains?
  • Macro implications: aggregate effects of scaling AI + capability building in MSMEs on employment, productivity, decarbonization trajectories.

Bottom line: This study suggests that AI raises firm‑level economic, environmental and social performance most effectively when paired with investments in organizational capabilities and sustainable business‑model innovation — a crucial consideration for economic models, policy, and investment decisions regarding AI diffusion in MSMEs.