Promoting the Adoption of AI-Based Recommendations Through Organizational Practices

Using artificial intelligence (AI), prescriptive process monitoring techniques suggest interventions to improve the efficiency of business processes and prevent negative case outcomes. These interventions aim to trigger process workers to adapt regular process execution in a specific case. Although this adaptation can aid process performance, process workers often do not react to them. The reasons for this reluctance are still opaque. Technical approaches in human-computer interaction try to increase the user’s attentiveness to interventions through prompts or seek to provide explanations for predictions by explainable AI (XAI). So far, these approaches have not sufficiently studied the relevance of the users’ organizational context and practices from a socio-technical perspective. This view helps us understand the influences on the willingness to react to system-based interventions. We conducted an analysis of research on prescriptive process monitoring and human-centered AI in organizations and explored an empirical case. By deriving 20 essential requirements, we designed a framework that represents a socio-technical meta-process of how AI-based recommendations could be organizationally embedded. For example, interventions can be amplified by co-workers, managers, and other stakeholders, explanations can be completed by human contribution, and reflection can be promoted by managers to trigger the evolution of AI. This framework can serve as a basis for further research on coordinating the users’ interactions with prescriptive process monitoring.