Gene regulatory networks (GRNs) are groups of genes that encode transcription factors responsible for gene expression changes during cell fate specification. One common GRN topology in lineage specification consists of two mutually repressive genes. A well-studied example of such a network involves GATA1 and PU.1 – two transcription factors that regulate the binary choice of myeloid progenitors to enter either the megakaryocyte-erythroid (ME) or the granulocyte-monocyte (GM) lineage. It has been proposed that random fluctuations of GATA1 and PU.1 determine cell fate, but more recent data has questioned whether these fluctuations are sufficient and have highlighted a role for cell-cell signalling. In this Issue, Megan Rommelfanger and Adam MacLean use a multiscale model that combines a stochastic cell-cell communication motif (described by a Poisson process) upstream of the deterministic dynamics of the GATA1/PU.1 GRN. Using the model, the authors record the impact of altering the mode of signalling (positive or negative feedback), the strength of cell connections and the spatial organisation of connections. In all cases, the cells reach a bistable state (ME or GM fate), but minor changes in these parameters have pronounced effects on the cell fate outcomes of the population. This model explains how the addition of cell-cell communication changes cell fate decisions from deterministic to probabilistic, and how a homogenous population can be transformed into a heterogenous bistable population, in line with experimental data. Together, these data highlight the importance of cell extrinsic signals in lineage specification.