Microsoft AI diagnostic system outperforms generalist doctors in complex case study

Traders Union
All News
Business & Wealth
Companies
Microsoft AI outperforms doctors

Highlights

Microsoft's AI diagnostic system DxO correctly diagnosed 83% of complex New England Journal of Medicine cases versus 36% for generalist doctors using reference materials.
Testing with multiple models found OpenAI's GPT-5 and GPT-5.5 delivered the best performance, with Microsoft's orchestration layer providing additional accuracy gains.
Microsoft plans real-world clinical trials with Mayo Clinic and aims for deep integration of DxO into health systems, potentially positioning it ahead of Google and OpenAI in healthcare AI.

Study results and healthcare AI roadmap

As reported by TechCrunch, citing Science, Microsoft’s AI diagnostic system correctly diagnosed 83% of complex medical cases in a simulation based on cases from the New England Journal of Medicine, compared with 36% for primary care doctors and general internists working individually with reference materials.

MAI-DxO, also called DxO, uses several AI agents designed to function like a panel of doctors. The system questions a patient, orders tests and debates possible diagnoses before producing an answer. Microsoft AI CEO Mustafa Suleyman describes the findings as a step toward what he calls “medical superintelligence.”

Microsoft tested the orchestration system with models from OpenAI, Google, Anthropic, Meta and DeepSeek, with the strongest results coming from OpenAI’s GPT-5 and GPT-5.5. The study data suggests the largest share of improvement comes from the underlying models themselves, while Microsoft’s orchestration layer adds further gains.

Suleyman says Microsoft is also working with Mayo Clinic to train a healthcare foundation model planned for release next year. He says DxO forms part of the infrastructure the company is building to pursue larger advances in medical reasoning and care coordination.

Clinical evaluation and competitive implications

Microsoft is now preparing real-world evaluations of the technology, according to Dominic King, vice president of Health at Microsoft AI. In the planned setup, patients will continue to receive standard care, including triage, specialist time, treatment and testing, while consenting patients also interact with DxO in parallel so its recommendations can be compared with those of human clinicians.

King says the wider healthcare challenge is a growing mismatch between demand for care and the supply of medical expertise. That framing positions the system not only as a benchmark exercise, but as a potential operational tool for hospitals and health systems facing staffing pressure and rising case complexity.

Microsoft’s publication places it alongside Google and OpenAI in arguing that advanced AI can match or exceed doctors on selected tasks. The comparison in this study is limited, however, because the difficult cases would typically be managed by specialists or physician teams, who were not included in the benchmark.

The company says its broader aim is to move beyond narrow tests such as medical exam questions and diagnosis generation toward deeper integration with clinicians and healthcare systems. Suleyman says that ambition could become a strategic differentiator for Microsoft as frontier AI companies compete to apply their models in high-value industries.

In our earlier article on AI agents in large U.S. companies, we reviewed survey findings showing many CEOs are seeing incremental efficiency and growth gains, but few consider the technology transformational yet. The piece highlighted that realizing bigger impact typically requires process redesign, data readiness, and continued hiring, retraining, and infrastructure investment rather than “plug-and-play” deployment.

This material may contain third-party opinions, none of the data and information on this webpage constitutes investment advice according to our Disclaimer. While we adhere to strict Editorial Integrity, this post may contain references to products from our partners.