Generative AI Literacy in Dutch Public Organizations: Competency Requirements for Responsible Governance and Use

Abstract

The rapid rise of generative artificial intelligence (AI) is transforming public organizations, raising questions about responsible use, governance, and workforce preparedness. While AI literacy has been widely studied in educational and technical domains, little is known about the specific competencies required for generative AI literacy in the public sector. This thesis addresses this gap by examining which competencies are needed in Dutch public organizations and how they vary across employee roles and organizational types. A qualitative research design was employed, combining nineteen semi-structured interviews with policymaking staff and developers across ministries and executive agencies with an additional analysis of government policy documents. A literature-based framework of generative AI competencies served as the starting point and was refined through abductive coding using Gioia-inspired structures. Competency requirements were systematically compared across two roles, policymaking staff and developers, and two organizational types, policy-based and service-based organizations. Findings reveal a shared foundation of competencies across all roles and organizations types, regarding basic literacy, critical assessment, ethical and legal implications, and prompting. Beyond this baseline, the emphasis diverges. Developers require deeper technical expertise in programming, finetuning, and system integration, whereas policymaking staff focus on governance, compliance, and boundary-setting. Similarly, service-based organizations emphasize experimentation, prototyping, and tool use, while policy-based organizations prioritize translating external regulations into internal rules and accountability structures. Findings also detail that generative AI literacy is not just an inventory of individual competencies but is also shaped by conditions such as policy clarity, access to technology, and organizational scope. These conditions influence the competency development within Dutch public organizations, which practices remain fragmented. Central government policies stress the importance of generative AI literacy, but access to training and secure tools is limited and unevenly distributed, creating a gap between formal ambitions and organizational practice. Overall, this research shows that generative AI literacy in the public sector cannot only be viewed a set of individual competencies, but also as an interdependent and context-dependent phenomenon. This interdependence stems from the complementary competencies of policymaking staff and developers, contextdependency is revealed by the identified conditions that shape the competency development. These insights highlight that fostering generative AI literacy requires more than individual training. It calls for an organizational environment that enables basic training to build a shared foundation, then role- and context-specific learning pathways and cross-role collaboration that allow employees to responsibly use and govern generative AI in service of the public good.