With recent improvements in Neural Machine Translation (NMT), Language Service Providers (LSPs) are challenged to embrace the NMT culture and become 'technology coaches' in order to support Enterprise Customers, manage translators and colleagues' expectations whilst getting accustomed to how NMT is changing the Localization Industry. From my experience working in a busy global LSP, I've been able to see people's reactions to deploying MT and decided to outline some common misinterpretations around its implementation. In this short talk, I'd like to encourage a conversation around questions such as: is doing MT really worthwhile if I don’t increase my volume for new words? Is training an engine going to solve all my problems? If I train the engine only with terminology, will I get a better MT output? Will the engine learn over time? Is implementing MT going to take a lot of time? (And what is "a lot"?). Even though these questions can be tackled during discovery conversations with Enterprise Customers, the wide-ranging information around MT in the Loc Industry (new features like the promising Quality Estimation, Adaptive MT, new support for low-resource languages, MT metrics, MTPE pricing, and the list is long) might end up being simply too much for consumption, leading to misinterpretations on its use. MT implementation is a process, beginning with choosing the right content for MT, but what if we want to go with "Full MT"? Is doing MT on every content even possible? The world of business is craving for a more widespread, clear distribution on MT education as there's still a long way to standardization.