Developing Competences for Artificial Intelligence in the Public Sector: The AI4Gov Canvas

This article introduces the AI4Gov Canvas, an evidence-based framework designed to identify and develop the competences needed by individuals working with artificial intelligence (AI) in the public sector. Recognising the need for functional specialist profiles in AI management, use and implementation within public administrations, we address the challenge of combining a variety of domains, such as technology, management and design, and policy, ethical and legal expertise. The AI4Gov Canvas was developed through a three-step process of competence identification, clustering, and validation, involving a review of research and grey literature, expert interviews (N=5), and an online survey (N=54). It is structured around three dimensions: technology, management and design, and policy/legal/ethical, with each dimension further categorized into three levels of abstraction: meta-competences, governance competences, and operational competences. The Canvas serves as a practical tool for mapping existing competences, identifying training needs, and designing functional specialist profiles that balance competence importance with the difficulty of obtaining them. We provide concrete examples of how the Canvas can be used to create personas, such as the "TechnoSteward" and the "Policy Sentinel", which highlight the framework's ability to guide the development of realistic and multi-disciplinary competence profiles. We conclude by outlining lessons learnt and future steps.