Prompting for reorganizing tacit Knowing: Epistemic practice in human-AI interaction

Prompting has rapidly become a central practice in human–AI interaction, yet it is often treated primarily as a technical procedure for controlling generative models. This article develops a sociomaterial epistemology of prompting in organizational and professional knowledge work, arguing that prompting can constitute an epistemic practice grounded in the partial externalization of tacit knowing. Drawing on Polanyi, Vygotskian mediation, sketch theory, distributed cognition, and boundary-object theory, the article conceptualizes prompting as a recursive process in which a user partially externalizes tacit knowing in a prompt, encounters LLM-generated linguistic variation, and interprets that material against existing understanding. Elements of the generated response may be resisted, appropriated, or partially internalized, thereby contributing to the reorganization of the tacit ground from which subsequent articulation emerges.This account differs from conversion-based epistemologies such as SECI: tacit knowing is not treated as content that can be transformed into explicit knowledge, but as a mode of attending that can only be partially articulated. Epistemic agency therefore remains human; the LLM contributes patterned linguistic variation that changes the material available for interpretation without itself knowing or interpreting. The article distinguishes three types of prompting within this recursive process: thinking prompts support individual exploration and partial externalization; talking prompts introduce socially distributed interpretation and negotiation across different tacit repertoires; and prescriptive prompts represent provisionally stabilized articulations that may circulate beyond the community in which their meaning was negotiated. The framework shows how prompting can reorganize tacit knowing while also enabling provisional stabilization and coordination in organizational knowledge work.