The advent of Neural Machine Translation (NMT) has raised quality expectations for MT output and further empowered stakeholders with repositories of data and the financial wherewithal to pay for expertise and training. This process has, in turn, sped up the polarization of the translation marketplace, with a focus on efficiency rather than quality at the lower end of the market. This presentation looks initially at NMT and data requirements, then more broadly at human factors in NMT, and at how automation of workflow steps is changing translation workflows.