AI solution adoption governance for SME

Abstract

With rapid evolution and diverse success stories of artificial intelligence (AI) technology, AI solutions are being developed and introduced in a variety of industries. However, due to the multifaceted considerations of adopting AI solutions, many companies are still hesitant to apply them or stuck at the proof-of-concept level. As with previous results in various research fields, AI adoption governance can be broadly classified into two choices: make or buy. In detail, the self-development based only on the company’s own resources or buying solutions already developed by other companies. AI solution adoption governance is closely related to Information System (IS) research fields because AI technologies are embedded in the Information Technology (IT) systems so that end users can easily use it. However, previous IS literatures are mainly focused on which benefit or how value can be created by adopting AI solutions and, as far as we know, do not particularly cover an AI solution adoption governance. This insufficient research trend is the main driver of this study. As multifaceted conflicting factors between costs and future capabilities need to be considered for companies looking to adopt AI solutions, transaction cost economies (TCE) and resource-based view (RBV) combined consideration is necessary. Even though a few IS research covers TCE and RBV combined decision making for innovation or outsourcing topic, but they treat only asset specificity as a major factor in TCE and recognize RBV is a more important decision-making factor. Besides IS fields, some papers simultaneously consider TCE and RBV, but they do not reflect AI solution related all factors. In this study, we suggest seven attributes based on two fundamental theories showing significant attention from previous governance related research. In TCE, four factors (asset specificity, technological uncertainty, behavioral uncertainty, and frequency) are used. Specifically, operation cost is considered through frequency factor which was often ignored n previous TCE literatures. This is because AI technologies cannot learn on their own, transaction costs according to the learning frequency should be assessed when selecting governance. From RBV, three factors (possibility to gain competitive power, absorptive capacity, and appropriability hazard) are used; We include various capabilities and knowledge in the category of resources by referring to the initial RBV definition. Absorptive capacity refers to the extent to which employees have prior knowledge of AI technology. It is reported as a key resource to properly utilize AI solutions. Appropriability hazard, the risk of the company’s unique competitive resources being leaked to the outside, are another consideration because companies are not only interested in acquiring the resources they need, but also protecting their own valuable, rare, inimitable, and non-substitutable (VRIN) resources. We limit sample data only small and medium-sized enterprises (SMEs) because large companies can be applying various adoption governance due to their relatively better financial and slack resources. Through conjoint experiments, we are planning to analyze which factors are most significantly considered for SMEs and check how the characteristics of SMEs moderate the main effect. In addition, we consider two things to increase the reliability of the experiment. First, a survey will be conducted on CEOs who are practical decision makers of SMEs. Second, redundant profile design is used to check consistency of respondents’ answers. This study is expected to contribute four aspects. First, it will identify significant factors affecting the AI solution adoption governance by synthesizing factors both related to the characteristics of the required resource adoption cost through TCE and each company’s idiosyncratic resource analysis through RBV. Second, the frequency factor overlooked in previous TCE literatures is used to determine governance