This chapter introduces cognitive capital and human-AI interactions. The upcoming cognitive revolution, like its industrial precursor, is defined by a radical change in technology. The machines of today introduce intangible cognitive capital that substitutes for human mental effort. The chapter lays out a three stage human-AI decision-making pipeline. This starts with human labeling of cases, continues with minimization of the AI loss function, and involves a final stage of human-AI collaboration. The chapter introduces outlines a field study in which elicited beliefs improve labeling for cancer detection. It introduces recent findings showing that standard engineering methods of defining loss functions can fail in their own terms. As a result, there is a bilateral alignment problem that requires not only that the AI is trained in human preferences, but that humans understand the AI's costs and constraints.