Microsoft Claims Its AI Diagnostician Outperforms Doctors in Complex Cases

Microsoft has unveiled its new AI-powered tool, the Medical AI Diagnostic Orchestrator (MAI-DxO), which the company says can diagnose complex medical cases much more accurately than human doctors. In a study using 304 challenging medical cases from the New England Journal of Medicine, the AI achieved an impressive 85.5% accuracy rate, while a group of 21 experienced doctors from the US and UK managed about 20% accuracy.

The system works by simulating the way real doctors approach diagnosis—asking follow-up questions, suggesting tests, and narrowing down possible conditions step by step. Unlike other single-model AI tools, MAI-DxO coordinates multiple advanced AI models, including OpenAI’s, Anthropic’s Claude, Google’s Gemini, and Meta’s Llama, functioning like a virtual panel of specialists.

Additionally, it plans diagnostic tests with cost-effectiveness in mind, resulting in an estimated 20–70% cost saving compared to standard approaches. Despite these promising results, the technology is still in its research phase and is not yet approved for real-world use. Microsoft aims to test it further through clinical trials and regulatory checks before any practical deployment.

The tech giant views this development as a move toward a future “medical superintelligence” that will work alongside human doctors—helping them diagnose patients more accurately and efficiently, rather than replacing them altogether.