Generative AI is moving from a peripheral tool to a core strategic partner in corporate governance, boosting decision quality when organizations cultivate shared human-AI intentionality and the right interface and trust structures; however, these gains are contingent on firm capabilities and contextual factors.

Main Finding

The paper documents the emergence of "Joint Agency" in corporate governance: generative AI (GenAI) and human leaders form shared-intentional partnerships that materially improve Strategic Decision Quality (SDQ). GenAI functions not merely as a support tool but as a mediating agent whose agency amplifies the benefits of human skill development and algorithmic trust. The effectiveness of AI-driven strategy depends critically on the human–AI interface and organizational conditions that enable collaborative intelligence.

Key Points

• Joint Agency defined: a human–machine shared intentionality in which GenAI contributes substantive, coordinated agency to high-level strategic decisions.
• GenAI as mediator: GenAI agency mediates the relationship between human skill development and algorithmic trust, meaning training humans and building trust increase AI’s effective agency, which in turn raises SDQ.
• Role shift: GenAI moves from peripheral analytic assistant to central "architect" of strategy, extending computational reach into complex, combinatorial strategic domains where it can outperform unaided humans.
• Human–AI interface matters: design features (explainability, interaction modality, feedback loops) and institutional practices (training, governance rules) determine whether GenAI’s agency translates into better decisions.
• Conditional benefits: Gains in SDQ are contingent — they require aligned incentives, skill complementarity, calibrated trust, and robust oversight to avoid overreliance or misalignment.
• Practical and theoretical contribution: Provides a framework for understanding AI as an active governance actor and for studying organizational adaptation to intelligent agents.

Data & Methods

• Conceptual mediation model: The core empirical claim is that GenAI agency mediates between (a) human skill development and (b) algorithmic trust, with SDQ as the outcome.
• Mixed-methods approach (typical in this literature and consistent with the paper’s claims):
  • Behavioral/field experiments or vignette studies simulating strategic tasks to measure SDQ under different human-AI interaction regimes.
  • Surveys and interviews of executives/managers to measure human skill development, perceived algorithmic trust, and interface characteristics.
  • Case studies of firms that have deployed GenAI into strategic functions to trace governance changes and outcomes.
  • Quantitative mediation analysis (e.g., structural equation models or causal mediation techniques) to estimate indirect effects of human skill/trust via measured GenAI agency on SDQ.
  • Robustness checks: alternative SDQ measures, heterogeneity by task complexity, and sensitivity to interface design features.
• Outcome metrics: Strategic Decision Quality (task performance, forecast accuracy, strategic ROI), measures of GenAI agency (degree of autonomous proposal generation, implementation influence), human skill indicators (training, experience), and trust indices.

Implications for AI Economics

• Firm-level productivity and returns:
  • GenAI as a strategic architect can raise firm TFP by improving SDQ in complex, high-value decisions; returns to AI investments are amplified by complementary investments in human capital and interface design.
  • Value creation from AI is endogenous to organizational capital — identical AI adoption can produce different economic returns depending on training, governance, and trust.
• Labor and skills:
  • Demand shifts toward higher-order strategic skills (overseeing, aligning, and integrating AI outputs) and interface-design competencies; routine strategic support roles may be displaced or upskilled.
  • Compensation and organizational roles should reflect joint agency responsibilities (e.g., hybrid human–AI accountability positions).
• Contracts, incentives, and governance:
  • Principal–agent models must be updated: principals may contract with hybrid human–machine teams rather than pure human agents; incentive schemes should consider AI influence and shared intentionality.
  • Board duties, fiduciary responsibilities, and audit practices need to account for AI’s architect role — transparency, verification, and accountability mechanisms become central.
• Market structure and competition:
  • Firms that successfully establish joint agency (skillful human complements + effective interfaces) may gain competitive advantages, potentially increasing market concentration in sectors where strategic complexity is high.
• Policy and regulation:
  • Regulators should focus on standards for explainability, interface safety, and governance processes rather than banning capabilities; policies that subsidize managerial training and interface development can accelerate socially beneficial adoption.
  • Antitrust and disclosure rules may need updating to capture AI’s role in strategic coordination and potential impacts on competition.
• Research directions for AI economics:
  • Quantify heterogeneity in returns to GenAI adoption across firms and tasks.
  • Model contract design and incentive provision when agents are hybrid human–AI teams.
  • Evaluate macro effects of widespread joint agency on investment, labor markets, and industry dynamics.

Overall, the paper reframes AI integration as an organizational and governance problem as much as a technological one: economic gains from GenAI hinge on institutions that cultivate complementary human skills, calibrated trust, and high-quality human–AI interfaces.