AI-Driven Leadership: Decision-Making, Competencies, and Ethical Challenges—A Systematic Review

Background: Artificial intelligence (AI) is transforming leadership and raising critical questions about decision-making, leadership capabilities, and ethical accountability in increasingly digitalized organizations. Objective: This systematic review synthesizes peer-reviewed evidence to answer: How does AI integration transform leadership and decision-making in organizations? Methods: A PRISMA 2020-compliant systematic review was conducted using structured Boolean searches in Scopus and Web of Science Core Collection on 26 February 2026. Eligibility was restricted to English-language, peer-reviewed, open-access journal articles with an explicit AI–leadership integration signal. Records were deduplicated and screened by two reviewers, with full-text assessment conducted against predefined criteria. A qualitative, narrative (conceptual) synthesis integrated heterogeneous empirical and conceptual contributions. Results: From 452 records, 84 studies met inclusion criteria. The synthesis identified three recurring analytical dimensions: (i) AI-augmented decision-making, (ii) leadership competencies and role shifts, and (iii) ethical challenges (accountability, transparency/opacity, fairness, privacy, and human agency). Integrating these dimensions, the review conceptualizes AI-driven leadership as a hybrid decision phenomenon in which AI accelerates and expands decision cycles, leaders reconfigure roles toward decision architecture and orchestration, and ethical conditions shape legitimacy, adoption, and authority dynamics. Conclusions: The review advances theory by specifying a mechanism-oriented model of AI-driven leadership and proposing testable propositions linking AI modality, role reconfiguration, and ethically conditioned legitimacy under key boundary conditions (e.g., sectoral stakes, governance capacity, and data/infrastructure readiness). Practically, it outlines an implementation pathway emphasizing decision criticality assessment, formalized human–AI task allocation, and institutionalized oversight mechanisms. Limitations: Findings are bounded by database selection and the open-access full-text constraint, which may under-represent paywalled scholarship.