Shadow Learning and Knowledge Redistribution: How Workers Appropriate Generative AI Outside Formal Organisational Systems

Abstract

ABSTRACT Organisations are ceding visibility over how their employees develop knowledge. As generative AI tools become widely accessible, workers are steadily building capabilities and co‐producing knowledge through informal AI engagements that operate entirely outside sanctioned training systems. This paper theorises shadow learning, the informal and autonomous appropriation of AI by workers for knowledge expansion beyond managerial design, as a process that fundamentally reconfigures how knowledge authority is distributed in organisations. Integrating critical management learning with sociomaterial and post‐human perspectives, and drawing on current debates in AI ethics and human–AI interaction, the paper conceptualises shadow learning through a recursive model of constraint, appropriation and reconfiguration. This framework shows how AI functions as a sociomaterial catalyst that alters who holds knowledge authority and how learning unfolds in practice. The paper makes three contributions. It introduces shadow learning as a concept that extends beyond informal learning by foregrounding its subversive and sociomaterial dimensions; it bridges critical and sociomaterial perspectives to show how AI‐mediated learning works at once as political resistance and distributed practice and it draws both traditions into contact with current debates in AI ethics. The contribution lies in that integration and its application to a phenomenon neither tradition has theorised alone, not in restating a post‐human case sociomaterial scholarship has already made. The model offers knowledge management scholars and practitioners a conceptual tool for understanding why informal AI‐mediated knowledge practices are not deviations from organisational learning systems but constitutive features of them.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 1 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.