Management knowledge diversity has a U-shaped effect on corporate AI innovation: medium diversity depresses AI innovation while very low or very high diversity boosts it; executive long-term orientation partly explains this pattern, and institutional investors and a developed digital economy tend to flatten the effect while cooperative innovation amplifies it.

Main Finding

Management knowledge diversity (the breadth of managers' functional/professional backgrounds across strategy, organization, HR, marketing, etc.) has a U-shaped relationship with corporate AI technological innovation. At low→medium levels, increasing management knowledge diversity depresses AI innovation (coordination costs, cognitive conflict); beyond a threshold, further diversity promotes AI innovation via information complementarity and cross-boundary integration. Executive long-term orientation partially mediates this U-shaped effect. Institutional ownership and regional digital-economy development weaken (flatten) the U-shaped relationship, while cooperative innovation strengthens (steepens) it. The effects are stronger in low-pollution industries, firms with high management shareholdings, eastern regions, highly competitive markets, and regions with stronger IP protection.

Key Points

• Theoretical framing: combines the knowledge-based view (knowledge as strategic, hard-to-imitate resource requiring integration) with temporal-orientation theory (executives' long-term orientation shapes risky, long-horizon investments).
• U-shaped mechanism:
  • Left/declining segment (low→medium diversity): narrow cognitive vision initially; as diversity grows coordination costs, communication barriers and cognitive conflict rise → AI innovation falls.
  • Right/rising segment (high diversity): once diversity passes a threshold, information complementarity and cross-boundary integration dominate → AI innovation rises.
• Mediation: management knowledge diversity → (U-shaped effect on) executive long-term orientation → higher AI innovation. Long-term orientation increases willingness to allocate resources to long-payoff, uncertain AI projects.
• Moderation:
  • Institutional ownership: monitoring/governance substitutes for some internal integration, reducing coordination costs in the declining segment but diluting marginal complementarity in the rising segment → flattens the U-curve.
  • Digital economy: mature digital infrastructure and platforms reduce internal communication/integration frictions, weakening the sensitivity of AI innovation to management knowledge diversity → flattens the U-curve.
  • Cooperative innovation (external partnerships): amplifies both costs at medium diversity and complementarities at high diversity → steepens the U-curve.
• Heterogeneity: promoting effect of high management knowledge diversity on AI innovation is more pronounced in specific contexts (low-pollution sectors, high management ownership, eastern China, competitive markets, strong IP protection).

Data & Methods

• Sample: 51,029 firm-year observations from 5,120 A-share listed Chinese firms, 2011–2024.
• Empirical strategy:
  • Hierarchical regression to test baseline relationships and interaction/moderation effects.
  • Instrumental variable (IV) approach to address endogeneity concerns.
  • Propensity score matching (PSM) as a robustness check for selection bias.
  • Heterogeneity analyses across industry, ownership, region, market competition and IP-protection subsamples.
• Variables: management knowledge diversity measured as breadth across management functional fields (non-technical); AI technological innovation measured (paper reports AI innovation outcomes — likely proxied by AI-related patents / investments, though precise operationalization should be checked in the full text).
• Robustness: multiple econometric approaches and subsample tests used to support the U-shaped finding and mediation/moderation results.

Implications for AI Economics

• For firm strategy and organization:
  • Management composition matters nonlinearly for AI capability. Firms should not assume monotonically positive returns from adding diverse managerial expertise; they must manage the integration process to cross the “coordination-cost” threshold where complementarities dominate.
  • Invest in mechanisms that reduce coordination costs (shared language, cross-functional processes, incentives for integration) to realize the upside of diversity sooner.
  • Promote executive temporal cognition (long-term orientation) — via incentive structures, governance norms, or leadership development — to translate diverse managerial knowledge into sustained AI investment.
• For investors and corporate governance:
  • Institutional investors can mitigate early-stage coordination frictions but may also substitute for internal integration benefits; stewardship strategies should balance near-term governance with support for internal knowledge integration.
  • High managerial ownership amplifies the positive effects of diversity in many contexts.
• For policy and regional development:
  • Digital-economy infrastructure can substitute for some internal integration needs and thus changes how firm-level managerial diversity affects AI outcomes; policymakers should recognize interaction effects between digital infrastructure and internal firm capabilities.
  • Policies that strengthen IP protection and encourage cooperative innovation can alter the shape and magnitude of diversity’s impact — cooperative innovation can magnify both risks and gains from managerial heterogeneity.
• For modeling and empirical work in AI economics:
  • Nonlinear micro-foundations matter. Models of firm-level AI adoption and innovation should allow for U-shaped effects of organizational attributes and include cognition/temporal-orientation mechanisms.
  • Future empirical work should distinguish management knowledge diversity from technological knowledge diversity (they have different integration logics and nonlinear patterns).
• Practical trade-offs: Encouraging managerial diversity is not a universal prescription; complementary investments in integration capabilities, governance design, and regionally appropriate digital infrastructure determine whether diversity yields net gains for AI innovation.

(Authors’ contributions include challenging the common inverted-U finding for technological knowledge by identifying a U-shaped pattern for management-knowledge diversity, and opening the micro-mechanism via executives’ long-term orientation. For precise variable definitions and operational measures—e.g., the exact AI-innovation and diversity metrics—see the full paper.)