Multi-componential models of translation competence are widely used in translator training as a yardstick for curricular and syllabus design. These models must be adapted to reflect professional trends, such as the impact of artificial intelligence, and machine translation in particular, on working methods. This paper describes the process of adapting a pioneering model of legal translation competence to the broader scope of institutional translation in light of recent trends, as verified by triangulating information from multiple interviews, analyses of translation volumes and job descriptors and other professional inputs. The resulting revised descriptor was validated through a survey of 474 translation professionals from 24 international organisations of diverse sizes and domain specialisations. The suitability of the descriptor was corroborated across the board, but variations were found in perceptions of the relevance of sub-competences to ensure translation quality. Profiles with a stronger specialisation in legal translation or more experience in institutional translation showed higher awareness of the relevance of all the sub-competences, especially the core language, strategic and thematic competences, and even more so for translating texts of a legal or administrative nature. The implications of these findings for training purposes in particular are discussed.