<p>Universities worldwide are facing a practical governance challenge as generative artificial intelligence tools become widely available to students and staff, while institutional responses remain fragmented across academic integrity rules, assessment design, research governance, and inconsistent understandings of acceptable assistance. This paper presents a pragmatic institutional toolkit for translating broad principles for ethical generative artificial intelligence use into workable higher education practice. Using a document-based design, the study draws on institutional evidence from a public higher education institution in Botswana, including governance documents, policy texts, role-specific guidelines, sensitisation plans, and workshop artefacts. The analysis shows that generative artificial intelligence governance can be operationalised without creating a standalone policy by embedding definitions and expectations within existing academic integrity and misconduct frameworks, developing separate guidance for different user groups, and using workshops, institutional memory and detection-tool training as educative and procedural mechanisms. The paper synthesises these elements into a roadmap-to-practice translation architecture, comprising five recurring translation moves, seven canonical toolkit components, and a six-step implementation pathway that emphasise transparency, proportionality, documentation, and due process. Rather than claiming effectiveness outcomes beyond the institutional record, the study offers a documented design case and an adaptable model for universities seeking coherent and feasible approaches to generative artificial intelligence governance, particularly in resource-constrained contexts.</p>