This study presents the results of a large-scale survey of institutional translation professionals on the distinctive features of post-editing (as compared to revision in particular), the main issues found in neural machine translation (NMT) and the impact of its use on translation processes, products and competences. The quantitative and qualitative findings illustrate the diversification of tools and inputs in current translation workflows, the risks and efforts associated with NMT, and the reinforced need for solid translation competence and subject matter knowledge to ensure translation quality. The (mostly limited) variations found between institutions and translators’ experience levels are also discussed. The evidence gathered supports a revised concept of translation that considers an increasingly hybrid and dynamic mix of translational actions in augmented translation environments. In turn, these results highlight the persistent relevance of translation expertise and specialized translator training, and contribute to nuancing our understanding of the implications of NMT for professional translation, beyond institutional settings.