Organelle targeting has emerged as a promising strategy in developing effective and specific cancer therapeutics by delivering a drug in its active form to the cellular compartment where it works. Among all the subcellular targets, endoplasmic reticulum (ER) targeting therapy has been little explored due to its complexity in cell signaling. As the largest cellular organelle, ER is responsible for crucial biosynthetic, sensing, and signaling functions in eukaryotic cells. Particularly, ER is responsible for the synthesis, folding, and posttranslational modifications of proteins destined for the secretory pathway, which amount to approximately 30% of the total proteome. Disturbing the protein-folding capacity of ER would result in ER stress, ultimately activating apoptotic signaling pathways and cell death. Therefore, selective disrupting ER function in cancer cells is a promising new strategy for anticancer therapies. However, current ER targeting small molecules, like tunicamycin and thapsigargin, lack cell selectivity and exhibit severe neurotoxicity, thus hindering their clinical applications. Therefore, it is necessary to develop novel ER targeting strategies that have high specificity against cancer cells. In this work, we employ enzyme-instructed self-assembly (EISA) to selectively target ER of cancer cells. Generated via enzymatic reactions on and in the cancer cells, the enzymatic assemblies interact with cellular membranes and disrupt plasma membrane integrity to enable the assemblies accumulate on the ER, thus inducing cancer cell death through ER stress. Utilizing enzymatic reactions and reduced diffusion of assembly, EISA enables spatiotemporal control of the generation and cellular distribution of the cytotoxic assemblies, thus providing a new strategy to regulate amyloid-like aggregates for treating cancer. This work, for the first time, demonstrates a reaction-based process for disrupting membranes in a spatiotemporally controlled manner, as well as subcellular organelle (i.e., ER) targeting, which illustrates a new concept in controlling cell fates via instructed-assembly.