With some form of AI perennially at the peak of the ‘hype cycle’, how are we finding useful applications for machine learning? Practitioners face many roadblocks, such as the availability of data, implicit bias, privacy considerations and overblown expectations. Transparency (or ‘explainable AI’) is seen as a particular challenge. These concerns are not unfounded. We present ways of addressing these concerns in a technology agnostic manner, with concrete examples.