Logistics firms reporting greater use of AI also report higher firm performance, largely through stronger innovation capabilities and improved logistics efficiency. The results are based on managers' perceptions in a cross-sectional Chinese sample and cannot establish causal effects or objective productivity gains.

Main Finding

Managers’ perceptions of logistics-oriented AI utilization are positively associated with perceived firm performance, and this relationship operates primarily through two complementary mechanisms — innovation capability and logistics efficiency. Both mechanisms produce significant indirect effects; a sequential path (capability → efficiency) is also significant but smaller, and the relative ordering between capability and process cannot be resolved with cross-sectional perceptual data. Managerial support did not moderate these relationships.

Key Points

• Theoretical framing: draws on the information-technology business-value perspective and dynamic capabilities theory to separate capability (innovation) and process (logistics efficiency) mechanisms linking AI use to performance.
• Sample: 254 middle- and senior-level managers in Chinese logistics firms; subsample of 194 involved in AI/digital-transformation activities for robustness checks.
• Main associations:
  • Perceived AI utilization → innovation capability (positive)
  • Perceived AI utilization → logistics efficiency (positive)
  • Perceived AI utilization → perceived firm performance (positive)
• Mediation results:
  • Both innovation capability and logistics efficiency produced significant individual indirect effects on perceived firm performance.
  • A sequential indirect path (AI → innovation capability → logistics efficiency → performance) was significant but smaller than each individual (parallel) indirect effect.
  • The two individual indirect effects did not differ significantly from each other, but both were larger than the sequential indirect effect.
• Model fit:
  • Sequential, reverse-sequence, and parallel-mediation specifications produced identical fit indices; a restricted direct-effects model showed weaker fit.
• Moderation:
  • Managerial support as a secondary boundary condition (moderator) did not yield significant interaction effects or moderated-mediation indices.
• Exploratory item-level findings:
  • Differentiated associations were observed for demand-forecasting and order-allocation applications and for AI infrastructure items.
  • The general pattern held in the AI/digital-transformation subsample.
• Limitations:
  • Cross-sectional, self-reported data — results reflect associations among managerial perceptions, not objective causal effects.
  • Sustainability/environmental outcomes not measured; implications limited to operational efficiency.

Data & Methods

• Data: Cross-sectional survey of 254 middle- and senior-level managers in Chinese logistics firms.
• Software/tools:
  • IBM SPSS Statistics 27 and IBM SPSS Amos 29
  • PROCESS macro v4.2 (Andrew F. Hayes), Model 83
• Inference: 5,000 bootstrap samples used to estimate indirect effects and confidence intervals.
• Analyses:
  • Tested parallel, sequential, and reverse-sequence mediation models involving perceived AI utilization → {innovation capability, logistics efficiency} → perceived firm performance.
  • Examined moderation by managerial support (interaction terms and moderated-mediation indices).
  • Conducted item-level exploratory analyses and robustness checks on 194 respondents involved in AI/digital-transformation activities.

Implications for AI Economics

• Mechanism differentiation: AI in logistics yields firm-level performance gains through two complementary channels — capability-building (innovation) and process improvement (logistics efficiency). Economic analyses of AI value should model both channels rather than aggregating effects into a single productivity parameter.
• Investment priorities: Because parallel (direct capability and process) pathways show stronger mediation than a capability→process chain, firms and policymakers should consider balancing investments in AI that directly improve operational processes (e.g., demand forecasting, order allocation, infrastructure) and those that expand innovation capability (new services, product/process innovation).
• Complementarity and substitution: The findings support complementarity between AI-driven innovation and operational efficiency; cost–benefit analyses and diffusion models should allow for joint production of capability and efficiency gains rather than treating them as mutually exclusive.
• Managerial factors: Managerial support did not significantly moderate effects in this perceptual sample, suggesting either (a) AI impacts are perceived to operate regardless of measured managerial backing, or (b) commonly used perceptual measures of support may not capture the relevant moderating mechanisms. Economic models that include managerial frictions should test multiple operationalizations and use objective adoption/usage indicators.
• Policy and evaluation: Because associations are based on perceptions and cross-sectional data, economic policy recommendations should be cautious. Objective, longitudinal data on AI deployments and realized firm-level outcomes (productivity, profits, employment, environmental externalities) are required to estimate causal returns to AI investments in logistics.
• Research directions: Future economic research should:
  • Use longitudinal or quasi-experimental designs to identify causal channels and ordering between capability and process effects.
  • Incorporate disaggregated AI applications (e.g., demand forecasting, order allocation) and infrastructure investments to estimate heterogeneous returns.
  • Measure objective performance outcomes (financial, productivity, environmental) to quantify welfare and distributional impacts of AI in logistics.