Microsoft AI Diagnostic Tool Outperforms Physicians in Study

Microsoft has developed the Microsoft AI Diagnostic Orchestrator (MAI-DxO), an AI-powered tool that diagnoses complex medical conditions more accurately than human generalist physicians. A study published in the journal Science found the system achieved 83% accuracy on complex cases from The New England Journal of Medicine, while generalists with reference materials reached 36% and those without reached 22%.

Microsoft built the system using multiple AI agents powered by OpenAI's GPT-5 and GPT-5.5 to simulate a medical panel. The tool mimics human clinical reasoning by sequentially questioning patients and ordering tests. In simulations at Beth Israel Deaconess Medical Center, the system demonstrated higher accuracy and lower estimated costs in emergency department settings.

Microsoft AI CEO Mustafa Suleyman described the tool as a step toward medical superintelligence. To validate these results, the company will begin real-world clinical testing at Beth Israel Deaconess Medical Center within the next few months. Additionally, Microsoft is partnering with the Mayo Clinic to train a healthcare-specific foundation model scheduled for release next year.