Automated Machine Learning – so-called AutoML —  has received considerable attention in recent years and is poised to take enterprise analytics to the next level. Most often, however, automation has been limited to the model-building algorithms themselves, such as hyper-parameter tuning and model ensembles. It appears that Insufficient progress has been made with the most time-consuming parts of the machine learning process: data preparation, model interpretation and model deployment. This talk will describe why attention in these steps has been slow in coming and practical recommendations for automating them.