This study examines how professionals perform boundary work while acquiring prompt engineering expertise for generative artificial intelligence (genAI). Based on 64 interviews with lawyers, knowledge managers, and technologists at two large law firms and a technology company, we identify three distinct learning pathways, each corresponding to relational expertise configurations and boundary work practices. Professionals engaging with genAI develop human-to-machine relational expertise, affording reinforcing boundary work. Those connecting with other professionals develop human-to-human relational expertise, affording relating boundary work. Those avoiding genAI do not develop relational expertise for prompt engineering, constraining them to resisting boundary work. Exceptionally, relational expertise from engaging or connecting learning pathways also affords resisting boundary work. We contribute a typology of prompt engineering expertise and a process model theorizing how professionals develop and apply it to redefine their work’s boundaries. By doing so, we link learning pathways, via relational expertise, to boundary work, providing insights into how expertise acquisition modes shape professionals’ agentic responses to emerging technology like genAI.
