Artificial Intelligence (AI) systems have demonstrated significant potential for advancement across various domains, including autonomous vehicles, intelligent personal assistants, and advanced robotics. Recent developments in generative AI have further highlighted this potential, particularly for knowledge-intensive tasks. However, growing public awareness of AI-related risks and the need to align AI systems with human and societal values has led to the development of ethical frameworks and regulatory measures. AI-specific regulations, alongside existing nondiscrimination and privacy laws, require AI governance in order to manage risks, ensure compliance, and uphold business ethics.

To address AI-specific governance challenges and promote transparency, fairness, non-maleficence, responsibility, and privacy, new governance tools and processes are required. Consequently, there is an increasing demand for empirical research on AI governance within organizations deploying AI systems. While information technology (IT) and data governance are established areas of information systems (IS) research, AI governance is yet an emerging field. The area contrasts with IT governance, as there are no existing governance models (such as Control Objectives for Information Technologies, COBIT) for AI.

This dissertation explores various organizational approaches to AI governance and examines how ethical principles and regulations are translated into strategic decisions, organizational processes, and practices. The dissertation comprises four articles. Article I is a systematic literature review analyzing 68 academic publications (out of 1071 identified) on organizational AI governance, elaborating on conceptual gaps in governance understanding and definitions. Additionally, Article I introduces key themes and future development areas for organizational AI governance. Article II provides an empirical perspective on how organizations translate ethical principles into practices, introducing four key translation practices. The research involved interviews with 13 frontrunner organizations deploying AI in their processes. Article III introduces a definition for AI governance in the organizational context and positions it within the broader landscape of corporate, IT, and data governance. Article IV examines contingency factors shaping AI governance approaches among organizations providing AI-assisted services in high-risk domains. It identifies seven contingency factors: volume of AI systems, industry sector, regulation, customer expectations, culture and values, strategic priorities, and technology and process maturity. The study also distills four archetypal AI governance approaches: differentiating, pragmatic, risk-taking, and disinterested.

Collectively, this research aims to provide theoretical and empirical insights on organizations translating ethical principles, regulations, and other external stakeholder pressure into AI governance practices. By offering definitions for AI governance, positioning it within a larger context, and introducing supporting frameworks, this dissertation contributes to the ongoing discussion on responsible AI and integrates with established research streams on IS planning, IT governance, and contingency theory.