What AI builds for organisations: an organisational capital perspective on the AI–HRM literature

Abstract

The literature around artificial intelligence (AI) and human resource management (HRM) has grown rapidly since 2019, yet it has theorised AI as a phenomenon acting on individuals rather than as a process that builds organisational structures. This matters because organisations making substantial investments in AI-driven HR systems currently lack any theoretical framework for evaluating whether those investments are accumulating as durable structural assets or dissipating as operational expenditure. This article identifies this structural gap by examining the extent to which the AI–HRM field engages with organisational capital (OC) theory, and proposes a new construct that addresses it. A theoretically anchored bibliometric analysis is conducted on 680 articles and reviews extracted from the Scopus database, spanning 1991–2026. The corpus was finalised through four sequential filters: subject area restricted to Business, Management and Accounting; document type restricted to articles and reviews; and language restricted to English. Performance analysis and science mapping techniques, including keyword co-occurrence analysis, thematic mapping, trend topic analysis and thematic evolution, are deployed using Biblioshiny and VOSviewer. OC theory is introduced as an interpretive lens applied post-bibliometric analysis to assess the presence or absence of asset-building constructs across the field's conceptual structure. The results show that constructs/themes related to OC, including structural capital, algorithmic routines, knowledge codification and firm-specific algorithmic assets, are systematically absent from the field's major conversation, co-occurrence clusters, thematic maps and both temporal periods of thematic evolution. The field shows a consistent individual-level bias that has deepened rather than corrected across 35 years of scholarship. Organisations deploying AI in HR functions need frameworks that help them evaluate whether AI adoption is generating durable structural assets or merely automating processes, and whether it is accumulating assets or structural liabilities through the codification of biased knowledge. The construct proposed here provides a theoretical foundation for both evaluations. This article introduces a new theoretical construct for the AI–HRM field, termed as algorithmic OC, which is defined as the codified, institutionalised and organisationally owned knowledge embedded in AI-driven HR systems that persists independently of the individuals who designed or operate them. The article establishes the bibliometric evidence base justifying its development and explicitly distinguishes this construct from traditional knowledge management and existing AI capability frameworks.