The centaur advantage: How interpretable AI and human-AI collaboration creates co-specialized resource bundles that can be a source for a competitive advantage

Artificial Intelligence (AI) models can either replace human decision-makers (i.e., automation) or collaborate with them (i.e., augmentation) in human resource management (HRM) decisions. AI models vary in transparency, ranging from black box over explainable to interpretable. However, we lack a clear understanding of how different combinations of AI-based decision-making approaches (i.e., automation, augmentation) and AI model types (i.e., black box, explainable, and interpretable) influence stakeholder (i.e., job applicants, employees, managers) acceptance or contribute to an organization’s competitive advantage through co-specialized resource bundles. Using the stakeholder resource-based view, I argue the automated approach consistently triggers negative stakeholder reactions, regardless of the AI model type. Although the augmented approach seems promising, not all human-AI model collaborations are equally effective. I argue that only the “centaur approach,” where humans and AI models iteratively refine each other’s decisions, leads to stakeholder acceptance and builds effective co-specialized resource bundles. Such effective co-specialized resource bundles create synergy among high-quality human capital, a motivated workforce, and emergent capital (i.e., the enhanced decision-making capabilities from sustained interaction between the AI model and stakeholders) that are more valuable than those created by the human-only decision-making approach. However, these dynamics only take place if the AI model is interpretable. Black box and even explainable AI models undermine stakeholder acceptance, the development of co-specialized resource bundles, and organizational performance.