Researchers at Germany’s German Cancer Research Center, the HI-STEM Stem Cell Institute, and Heidelberg University Hospital have developed HemaGuide, a clinical artificial intelligence assistant designed to streamline treatment decisions for complex blood cancers. Published in Nature Medicine, the system addresses the growing therapeutic complexity of hematological oncology, where physicians must simultaneously evaluate patient histories, genetic mutations, comorbidities, and an expanding drug portfolio. Traditionally resolved through resource-intensive interdisciplinary tumor boards, these decisions are now supported by HemaGuide, which operates as an automated digital clinical conference. HemaGuide processes unstructured physician notes, synthesizes them with current medical guidelines, and cross-references a proprietary repository of over two thousand real-world tumor board cases alongside recent scientific literature. The architecture generates transparent, evidence-based treatment recommendations. A core capability is its integrated molecular tumor board functionality. The system identifies clinically relevant genetic alterations in under one minute, evaluates them against international oncology standards, and suggests targeted therapies. This rapid molecular assessment replaces a workflow that historically demanded hours of specialist time and was restricted to a handful of major centers. Clinical validation demonstrates robust performance across diverse hematological malignancies. In an evaluation of five hundred fifty-five independent tumor board cases spanning forty-seven distinct blood cancer subtypes, HemaGuide recommendations aligned with expert panel decisions in nearly eighty-two percent of instances. A concurrent prospective trial phase yielded a comparable agreement rate of approximately eighty-three percent. The system significantly outperformed conventional large language models when assessed on highly complex cases, particularly regarding patient-specific contextualization. Furthermore, junior physicians utilizing the assistant achieved decision-making proficiency approaching that of senior hematologists, with zero instances of misclassifying carcinogenic mutations as benign, ensuring a high safety threshold. Engineered for secure institutional deployment, HemaGuide runs entirely on local hospital servers, guaranteeing that sensitive patient data never leaves institutional infrastructure. Development leaders emphasize the tool functions strictly as a decision-support mechanism, augmenting rather than replacing clinical expertise. By accelerating molecular evaluations and standardizing complex assessments, the system aims to democratize access to specialized hematology care, particularly for regional hospitals lacking dedicated subspecialists. A multi-center clinical trial to assess long-term patient outcomes and healthcare quality improvements is currently in preparation.