Breast cancer remains a leading cause of death worldwide. Although chemotherapy reduces primary and metastatic tumour burden, persisting drug-tolerant tumour cell populations, known as minimal residual disease (MRD), pose a significant risk of recurrence and therapy resistance. In this study, we describe the spatiotemporal organisation of therapy response and MRD in BRCA1;p53-deficient mouse mammary tumours and human clinical samples. By integrating single-cell RNA sequencing, spatial transcriptomics, and imaging mass cytometry across multiple treatment timepoints, we characterise dynamic interactions between tumour cell subpopulations and their surrounding microenvironment. Our multiomic analysis uncovers a distinct, chemotherapy-tolerant epithelial-mesenchymal transition (EMT) cancer cell population that displays a conserved expression programme in human BRCA1-deficient tumours, significantly correlates with adverse clinical outcomes, and can be pharmacologically targeted in preclinical models. We reveal the spatial distribution of residual EMT-like tumour cells within discrete anatomical niches, providing a framework for understanding the persistence of MRD and potential therapeutic vulnerabilities.