Firms that build AI-enabled dynamic capabilities report better performance largely because AI fosters a data-driven mindset that institutionalizes learning; however, too much reliance on data weakens managers' flexibility and reduces marginal gains.

Main Finding

AI-enabled dynamic capability (AIDC) improves firm performance both directly and indirectly by fostering a data-driven culture (DDC) and organizational learning (OL). However, DDC has a nonlinear, boundary role: up to a point it enables AI value creation, but excessive reliance on data weakens the marginal performance benefits of AIDC by constraining managerial discretion and strategic flexibility.

Key Points

• The study integrates Resource-Based View (RBV) and Dynamic Capabilities Theory (DCT) to explain why AI investment does not automatically yield superior firm performance.
• AIDC has a significant direct positive effect on firm performance.
• Indirect paths:
  • AIDC → Data-Driven Culture (DDC) → Firm Performance (mediating effect).
  • AIDC → Organizational Learning (OL) → Firm Performance (mediating effect).
  • AIDC → DDC → OL → Firm Performance (serial mediation): AIDC builds a data mindset that then institutionalizes learning processes to convert AI-derived insights into routinized action.
• Moderation: DDC moderates the AIDC→performance relationship nonlinearly (threshold/inverted-U pattern). Moderate DDC amplifies AIDC benefits; excessive DDC weakens them, revealing a “dark side” of over-reliance on data.
• Managerially, firms must balance analytical rigor with adaptive learning and preserve managerial discretion/strategic flexibility to sustain AI-driven performance gains.

Data & Methods

• Sample: 254 senior managers and executives from U.S. firms actively using AI technologies.
• Design: Cross-sectional survey of firm-level perceptions (senior managers/executives as respondents).
• Analytical approach: Partial Least Squares Structural Equation Modeling (PLS-SEM).
  • Tested direct effects, parallel and serial mediation (DDC and OL), and nonlinear moderation (DDC as a contingent moderator with diminishing returns).
• Robustness/limitations noted by study (implied by method):
  • Cross-sectional survey limits causal claims.
  • Self-reported firm performance and single-source data may introduce common-method bias.
  • Sample limited to U.S. firms and senior managerial respondents—generalizability may be constrained.

Implications for AI Economics

• Returns to AI are conditional: Economic value from AI depends critically on complementary organizational capabilities (dynamic capabilities, data culture, learning processes), not just technology input.
• Complementarities and complementarities’ limits:
  • Investment complementarities: Organizational capital (learning systems, managerial practices) complements AI capital and determines realized productivity gains.
  • Nonlinear effects: There can be diminishing or negative marginal returns to intensifying a data-driven culture, implying an optimal intensity rather than “more is always better.”
• Allocation and adoption:
  • Firms and investors should budget for complementary organizational investments (training, learning processes, governance) when evaluating AI projects.
  • Policymakers and economists studying AI-driven productivity should account for heterogeneity in organizational capabilities and threshold effects in data reliance when estimating aggregate effects of AI diffusion.
• Measurement and research recommendations:
  • Future empirical work should use longitudinal designs and objective performance metrics to better identify causal effects and dynamic processes.
  • Model heterogeneity by industry, firm size, and AI application type to establish where DDC thresholds and returns to AIDC vary.
• Managerial practice:
  • Optimize the balance between analytical discipline and strategic adaptability: build data literacy and learning routines but retain managerial judgment and flexibility to avoid rigidity that reduces dynamic reconfiguration and innovation.