Responsible AI governance in practice: The strategic impact of responsible AI governance on business value and competitiveness

Abstract

Artificial Intelligence (AI) comprises a set of technologies with vast potential applications, which can range from autonomous vehicles and chatbots to fraud detection and medical diagnosis. While the domain of AI research has a history of over 50 years, recent technological advances have facilitated their utilization and deployment in real-world applications. Despite growing rates of AI deployment, many organizations struggle to fully realize business value from such technologies. Additionally, although AI offers numerous advantages, it is not exempt from potential negative or unintended consequences. Rising concerns regarding AI usage and instances of failed AI applications, some of which resulted in fatalities, job displacement, or racial biases, have underscored the urgent need for responsible AI governance (RAIG). Therefore, a gap exists between the design, deployment, and use of AI and its business value.

The thesis employs a sequential multiple methods research design, commencing with an exploratory approach to uncovering key aspects of RAIG and responsible principles. Initially, a comprehensive literature review was conducted to acquire a holistic understanding of AI use and its business value. A second literature review followed, exploring RAIG, aimed at grasping the principles and methods for ensuring responsible AI utilization. Following the two literature reviews, we conducted a series of empirical studies. We start with a multi-case study with three organizations to examine how responsible AI governance is implemented in practice and promotes the development of robust AI applications that do not introduce negative effects. Next, an in-depth case study with 14 expert interviews was conducted to explore the importance of RAIG and the dark side effects that might occur if RAIG is not present. After that, research was carried out to construct a conceptual framework, which forms the main processes involved in AI resource orchestration. This framework aims to explain the different activities used to orchestrate resources strategically, thereby generating business value. Finally, a quantitative study with 300 responses from Europe and the USA was undertaken to investigate whether RAIG yields tangible value and, if so, through which mechanisms and processes this value is realized.

The results contribute to our understanding of how RAIG is implemented in organizations, and what its resulting business value is: firstly, through a conceptual model by exploring the fundamental dimensions relevant to RAIG within organizations and unveiling the underlying practices supporting them; secondly, by identifying the negative or unintended consequences of AI in the absence of RAIG, categorized into three clusters related to the nature of work, conflicts and effects, and responsibility; and thirdly, through a conceptual model by presenting and elucidating how firms manage their RAIG practices to improve competitiveness. Finally, the research discusses implications for research, practice, and policy, while also highlighting avenues for future investigation.