In the public sector, AI is increasingly being adopted and implemented by organisations across different domains including education, healthcare, social services, policymaking, etc., to modernise and improve public service access and delivery and create greater public trust between government and citizens. Amidst its promise of delivering more efficient and equitable public services, the governance of AI - understanding and mitigating its risks and maximising its potential benefits - continues to generate and attract mainstream debate and attention in academia and amongst policymakers, leading to the development of different governance approaches including regulations like the EU AI Act and soft law governance which advocates and emphasises the application of various ethics principles, guidelines, etc., as a means of effectively governing the technology. However, there is limited empirical research on AI Governance (AIG), particularly on how these high-level ethical principles are operationalised in practice. My research investigates AIG practices in the public sector to fill this gap. The study applies a qualitative research methodology. Drawing primarily on empirical data collected via in-depth semi-structured interviews from Irish public sector bodies, the study adopts the Institutional Logics (IL) perspective to fulfil three main objectives: (1) explore the contextual factors shaping the adoption, implementation and governance of AI in the Irish public sector; (2) investigate how ethical principles are interpreted and translated into governance practices at the organisational level by agencies adopting and deploying AI; and (3) identify the societal-level and field-level logics underpinning the context and governance practices (in 1 and 2 above). The institutional logics perspective offers researchers a robust analytical and theoretical framework for empirical analysis (Cai and Mountford, 2022).