Microsoft Expands AI Health Business Through Doctor Trials

Microsoft’s MAI-DxO diagnosed complex cases correctly 83% of the time in a simulation, compared with 36% for primary care doctors and general internists working individually with reference materials, according to a study published Thursday in Science. The comparison used cases from the New England Journal of Medicine. It didn’t include specialists or physician teams, who usually handle cases of that complexity.

Microsoft sells dictation and radiology tools to clinicians and describes DxO as an expansion of those products. Mustafa Suleyman, Microsoft AI’s CEO, called the result a step toward “medical superintelligence.” The study measures diagnostic accuracy and estimated operating costs, while Microsoft’s plans link the system to new clinical evaluations and a healthcare foundation model.

Who Controls the System

MAI-DxO, also called DxO, coordinates several AI agents that question patients, order tests and debate diagnoses. Microsoft tested it with models from OpenAI, Google, Anthropic, Meta and DeepSeek. Its best results came with OpenAI’s GPT-5 and GPT-5.5.

The study data indicated that much of the improvement came from more powerful underlying models, with Microsoft’s orchestration system adding gains. Microsoft researchers said accuracy gains naturally shrink as models approach the theoretical maximum. A spokesperson said, “The DxO orchestration helps with the broader challenge of optimization both accuracy and cost.”

With a short, basic prompt, GPT-5 scored 78.9% on complex cases at about $8,400 per case. DxO cut the estimated cost by about $3,000 and raised accuracy by about four percentage points. On emergency department cases, GPT-5 with and without DxO showed little improvement in accuracy or cost.

GPT-5.5, released in April 2026, scored 72% on complex cases on its own; with DxO, accuracy rose to nearly 87% at a slightly lower cost. On emergency department cases, both versions scored about 85% at similar cost. Gemini 2.5 Flash scored 46.4% on its own and nearly 69% with DxO. DxO helped most on complex cases and with less capable models, though the performance gap between weaker and stronger models exceeded the boost from DxO.

Clinicians and the Commercial Push

Dominic King, vice president of Health at Microsoft AI, told TechCrunch, “The biggest challenge in healthcare delivery at the moment is the growing mismatch between demand and supply of medical expertise.” He said Microsoft plans to evaluate DxO’s clinical reasoning across hospital settings through a new partnership with Beth Israel Deaconess Medical Center in Boston.

Patients in the study will receive standard treatment, including triage, time with a specialist, treatments, investigations and a decision on whether it’s safe to discharge. Patients who consent will also engage with DxO. Microsoft and its partners will compare the system’s recommendations with those of human doctors. King said the study will include emergency and urgent-care settings and provide evidence about whether the systems could be deployed at greater scale in the future.

Microsoft is also working with the Mayo Clinic to train a healthcare foundation model, with release planned for next year, Suleyman said. He described DxO as part of Microsoft’s foundational infrastructure for healthcare breakthroughs. “We have to have the absolute number one state-of-the-art medical clinical decision support products over the next few years,” Suleyman said.

Claims Await Real-World Testing

King said the technology is ready for real-world evaluations. The planned study will compare DxO’s recommendations with doctors’ recommendations while patients continue to receive standard treatment. Microsoft’s claims of improved performance in complex cases come from a simulation, while the Boston partnership aims to assess the system in hospital settings.

Google published a clinical study with Beth Israel in The Lancet this month to evaluate its AMIE conversational diagnostic AI during real-world urgent-care visits, with human physicians monitoring interactions in real time. Microsoft said its study would be the biggest of its kind.

Suleyman said generative AI’s use in healthcare remains at an early stage and predicted healthcare models would become “an order of magnitude better in capability within the next 24 to 36 months.” He said Microsoft expects to use real-world usage to identify gaps in models and improve them, describing that process as a “flywheel.” He also said the company expects a coding-style breakthrough for medical records in 2027 and wants to integrate AI with clinicians and healthcare systems to improve outcomes.