Chapter 2 is about automated decision-making, which is one area where AI, specifically predictive AI, is increasingly used: predicting who will commit a crime, who will drop out of school, and so forth. We’ll look at many examples of systems that have failed and caused great harm. In our research, we’ve identified a recurring set of reasons these failures keep happening—reasons that are intrinsic to the use of predictive logic in these high-impact systems. We’ll end the chapter by asking if it is possible to reimagine decision-making without predictive AI, and we’ll discuss what sorts of organizational and cultural adaptations we’ll need in order to embrace the unpredictability inherent to consequential decisions.