Nature Reviews Clinical Oncology, Published online: 07 August 2026; doi:10.1038/s41571-026-01189-0Oncology clinical trials are characterized by high failure rates — reflecting both the biological complexity of cancer and operational inefficiencies — and the generalizability of their results is often limited. In this Review, the authors discuss opportunities to address these limitations via application of artificial intelligence tools across the lifecycle of clinical trials as well as key challenges that must be overcome for routine clinical deployment of these tools.