Altered capacity for self‐renewal and differentiation is a hallmark of cancer, and many tumors are composed of cells with a developmentally immature phenotype. Among the malignancies where processes that govern cell fate decisions have been studied most extensively is acute myeloid leukemia (AML), a disease characterized by the presence of large numbers of “blasts” that resemble myeloid progenitors. Classically, the defining properties of AML cells were said to be aberrant self‐renewal and a block of differentiation, and the term “differentiation therapy” was coined to describe drugs that promote the maturation of leukemic blasts. Notionally however, the simplistic view that such agents “unblock” differentiation is at odds with the cancer stem cell (CSC) hypothesis that posits that tumors are hierarchically organized and that CSCs, which underpin cancer growth, retain the capacity to progress to a developmentally more mature state. Herein, we will review recent developments that are providing unprecedented insights into non‐genetic heterogeneity both at steady state and in response to treatment, and propose a new conceptual framework for therapies that aim to alter cell fate decisions in cancer. This review by A. Lewis and L. Kats summarizes the recent advances in our understanding of "differentiation therapy". The authors share new developments into non‐genetic heterogeneity, and propose a conceptual framework for therapies that aim to alter cell fate decisions in cancer.