Utilising Artificial Intelligence to Enhance the Accessibility of Records with Enduring Value in South African State-Owned Companies

Abstract South African state-owned companies (SOCs) hold records of enduring value - records that matter for accountability, for service delivery, and for the national documentary heritage. Yet in the digital era these records are increasingly scattered across shared drives, departmental file shares, email systems, and personal devices, where records managers and archivists cannot see them. The paradox is that SOCs now generate more records of enduring value than at any point in their history, while the proportion that can be discovered, appraised, and transferred to archival custody has declined sharply. Drawing on doctoral research conducted at the University of South Africa, together with the author's implementation experience with open-source archival and records management platforms, this paper argues that artificial intelligence (AI), applied within a socio-technical framework that retains human oversight and standards compliance, offers a practical response to this accessibility crisis. It presents a theoretical framework for AI integration in records management, summarises empirical findings from a qualitative single case study with records managers and technical experts at a South African SOC, and then sets out, in practitioner-accessible terms, how the framework can be implemented across the records lifecycle. The empirical study did not test a deployed system; the findings report participants' situated expert perspectives and design projections, including a projected transition from a 6.5% baseline discoverability rate to 95%+ within an 18-month phased rollout. The paper closes with recommendations for the responsible adoption of AI in South African archival and records management practice, grounded in the National Archives and Records Service of South Africa Act, the Protection of Personal Information Act, and the Promotion of Access to Information Act.