Managers’ trust, not just AI capability, determines whether AI improves strategic decisions; in highly complex tasks trust translates into smaller gains, suggesting staged human–AI workflows rather than full delegation.

Main Finding

AI Technical Competence (AITC) increases managers’ Algorithmic Trust, and that trust fully mediates the effect of AITC on Strategic Decision Quality. However, Task Complexity weakens the positive effect of Algorithmic Trust on decision quality: in hyper-complex tasks, trust translates less effectively into better strategic decisions. The net implication is that technical capability alone does not guarantee superior strategic outcomes—managerial trust and task context determine whether AI capability produces value.

Key Points

• Framing: AI is treated as an information-processing capacity within Organizational Information Processing Theory (OIPT); human Algorithmic Trust functions as the cognitive processor that links capacity to outcomes.
• Mediation: Algorithmic Trust fully mediates the relationship between AI Technical Competence and Strategic Decision Quality — meaning competence affects decisions primarily via trust, not directly.
• Moderation: Task Complexity exerts a “dampening effect” — as complexity increases, the benefit of trust for decision quality decreases.
• Practical prescription: The authors recommend hybrid-sequential workflows (human + AI collaboration with staged handoffs) rather than full delegation to AI in complex strategic settings.
• Context: Study uses a sample of 365 managers in China, characterized by rapid digital adoption and serving as a “global digital laboratory.”

Data & Methods

• Sample: 365 managers based in China (cross-sectional survey).
• Theoretical lens: Organizational Information Processing Theory (OIPT).
• Key constructs:
  • AI Technical Competence (AITC) — AI capability/technical performance.
  • Algorithmic Trust — managers’ trust in AI outputs/algorithms.
  • Strategic Decision Quality — outcome measure of strategic decision performance.
  • Task Complexity — moderator capturing task/environmental complexity.
• Analytical approach: Partial Least Squares Structural Equation Modeling (PLS-SEM) to test a moderated-mediation model (AITC → Algorithmic Trust → Strategic Decision Quality; Task Complexity moderates Trust → Quality).
• Findings summary:
  • Significant positive path: AITC → Algorithmic Trust.
  • Full mediation: Algorithmic Trust accounts for the AITC → Strategic Decision Quality relationship.
  • Significant interaction: Task Complexity reduces the positive effect of Algorithmic Trust on decision quality.
• Methodological caveats (implicit/likely):
  • Cross-sectional survey limits causal inference.
  • Single-country sample (China) may affect external generalizability.
  • Reliance on perceptual/self-report measures (if used) could introduce bias.

Implications for AI Economics

• Rethinking returns to AI investments: Economic value from AI depends not just on technical capability but on managerial trust and task context. Cost–benefit models should include human trust and task complexity as modifiers of expected payoff.
• Adoption vs. value paradox: The “AI paradox” (widespread adoption without superior strategic outcomes) can be explained by missing trust and unsuitable task fit; models of diffusion and productivity should incorporate cognitive and workflow factors.
• Optimal task allocation and workflow design:
  • For routine/low-complexity tasks, higher Algorithmic Trust more reliably converts AI capability into decision quality — supports broader automation and delegation.
  • For high-complexity strategic tasks, hybrid-sequential workflows (staged human-AI collaboration, human-in-the-loop checkpoints) preserve decision quality better than full automation.
• Labor and organizational design:
  • Investments in upskilling, trust-building (transparent explanations, calibration), and managerial training can raise the realized ROI of AI.
  • Job designs should reflect complementarities: deploy AI where trust translates to quality, keep humans in roles where complexity undermines trust’s effectiveness.
• Policy and governance:
  • Regulators and firms should promote transparency, interpretability, and interfaces that support calibrated trust; metrics for AI impact evaluations should measure trust and task complexity, not only algorithmic accuracy.
  • Public procurement and subsidy models might prioritize hybrid designs and human-centered adoption pathways for complex strategic domains.
• Measurement and macro modeling:
  • Empirical models of productivity and growth should include mediating variables like Algorithmic Trust and interaction terms for task complexity to avoid overestimating AI’s direct contribution to economic outcomes.
• Research suggestions for AI economics:
  • Test external validity across countries/industries and longitudinally to assess causal dynamics.
  • Quantify economic losses/gains from miscalibrated trust under varying task complexities to inform investment and policy thresholds.

Actionable takeaway: To capture the economic value of AI, firms should invest not just in technical competence but in building calibrated algorithmic trust, and they should match workflow design to task complexity—favoring hybrid-sequential processes for strategic, high-complexity decisions.