GLMs have become a standard for ratemaking in many lines of business. Recently there has been much attention on machine learning tools in ratemaking as well. The message of how these two analytical approaches can work together to improve segmentation accuracy is not always clear. This session will begin with a high level overview of best practices in GLMs. We will also introduce a new development to detect and simplify interactions within a GLM. The session will then explore how machine learning can augment GLMs in terms of whittling down unmanageable lists of explanatory variables, detecting interactions, and harnessing the information contained in model residuals in a very controlled manner that is suitable for implementation.