This paper presents a typology of human-AI work configurations that extends the discourse on AI in the workplace beyond the conventional automation-augmentation dichotomy and a predominant focus on “white-collar” knowledge work. The typology identifies eight distinct human-AI configurations organized along three dimensions: the distribution of performance, the distribution of control over work, and the distribution of recognition for work. Each configuration is illustrated by examples from the literature and is characterized by its typical relational logic: human autonomy, masquerading automation, subordination, collaboration, co-optation, operation, face-saving, and (true) automation. This work thereby offers a more nuanced framework for understanding a broader spectrum of human-AI relations in the workplace. It serves researchers and policymakers to explicitly evaluate the transferability of their findings and proposals from one context to another. Different relationship types are governed by different logics, which lead to different implications for work conditions and thus beg different appreciations and interventions.
