What Did You Do Now? A Qualitative Case Study on Knowledge Worker' Interactions with Agentic AI

Sammanfattning

This study draws on 22 semi-structured interviews with knowledge workers using an AAI-powered web browser across diverse professional contexts (e.g., consulting, engineering, marketing, and operations) to explore how AAI reshapes goal-oriented action in knowledge work. Employing Affordance Theory as our theoretical lens, we show how workers navigate a complex landscape in which the same technological features can both enable and constrain goal attainment, depending on the situated alignment between workers' goals and technological capabilities. We identify four affordances (e.g., automating workflows) and four constraints (e.g., increasing idle time) that emerge from knowledge workers' interactions with the material technological features of AAI. Further, our findings indicate that the likelihood of actualizing action potentials is shaped by contextual factors at the individual (e.g., AAI literacy), task (e.g., stakes and scale), and organizational (e.g., organizational ways of working) levels. This study offers one of the first empirical investigations of autonomous AI agents in knowledge work and contributes an empirically derived conceptual model of knowledge worker-AAI interactions. This contributes to a broader understanding of how knowledge workers' interactions with autonomous technologies generate more fundamentally agentic, rather than purely informational, action potentials, and illustrates how knowledge workers not only interrogate what AAI generates but also actively assess whether to delegate agency for action responsibilities - creating new accountability dynamics in knowledge work.