"The Copilot that Can't Fly": Evolving Human-GenAI Configurations for Augmentation

With generative artificial intelligence (GenAI) assistants becoming increasingly embedded in platform-based ecosystems like Microsoft Office, researchers and practitioners alike praise their augmenting potential for human creativity and productivity. However, little is known about how individuals configure these systems into their work to leverage these promises. Based on 57 semi-structured interviews with 12 IT service professionals, we examine how individuals experiment with and adapt to Microsoft’s GenAI assistant Copilot for M365. We identify six human-GenAI configurations shaping interaction patterns, affordances, and perceived augmentation outcomes. Our findings reveal two configuration behaviors: While some maintain rigid expectations and remain limited to basic functionalities, others refine their strategies over time, uncovering new affordances that integrate Copilot more actively in their work. Based on these insights, we propose a theoretical model of reconfiguration work in the context of GenAI systems, highlighting the critical role of individual engagement in realizing GenAI’s augmentation potential.