When and how are professionals able (or unable) to elicit expertise from domain experts over whom they have no formal authority? We examine this research question by drawing on four years of qualitative field work, comparing two AI development projects —involving one successful and one unsuccessful attempt to elicit domain expertise—within a large multinational fashion company. We unpack the interplay between task and organizational structures in enabling (or constraining) the effectiveness of AI developers in eliciting domain expertise. In particular, we show that in situations that are characterized by jurisdictional clarity (versus ambiguity), task centrality (versus peripherality), and task enactment homogeneity (versus heterogeneity), AI developers were more effective in accessing domain experts and eliciting their expertise. Building on these findings, we develop a model outlining how the interplay between task and organizational structure shapes both the legibility of domain experts as well as the concentrated nature of domain expertise, and its consequences for the effective (or ineffective) elicitation of domain expertise.
