Splicing factor 3b subunit 1 (SF3B1) is the most commonly mutated RNA splicing factor identified in myelodysplastic syndrome (MDS), chronic lymphocytic leukemia, and uveal melanoma. The mechanisms by which SF3B1 mutations promote malignancy are poorly understood. Here, we integrated pan-cancer RNA sequencing to identify mutant SF3B1-dependent aberrant splicing events with a positive CRISPR screen to prioritize alterations that functionally promote oncogenesis. Our results indicated that diverse, recurrent SF3B1 mutations converge on the repression of bromodomain containing 9 (BRD9), a core component of the recently described non-canonical barrier-to-autointegration factor complex (ncBAF). Mutant SF3B1 recognizes intronic sequences within BRD9 as exons, thereby permitting inclusion of aberrant sequence (i.e., poison exon) that will result in the degradation of BRD9 mRNA. BRD9 depletion results in significant loss of ncBAF at CCCTC-binding factor (CTCF)-binding loci but has no impact on the localization of canonical BAF. These actions resulted in disturbed myeloid/erythroid differentiation and promoted the development of MDS and melanoma. Of note, correcting BRD9 mis-splicing in SF3B1-mutant cells with antisense oligonucleotides (ASOs), by targeting the poison exon with CRISPR-directed mutagenesis, or via the use of spliceosomal inhibitors are all potential therapeutic options. Our results implicate disruption of ncBAF as a critical factor promoting the development of the diverse array of cancers that carry SF3B1 mutations and suggest a mechanism-based therapeutic approach for treating these malignancies.