Microsoft says its health AI diagnoses patients 4x better than physicians, but will it replace doctors?

Microsoft says its new health-focused artificial intelligence system has outperformed human doctors in diagnosing some of the world’s toughest medical cases, achieving accuracy more than four times higher than physicians in recent tests.

In a series of posts on Monday, Mustafa Suleyman, CEO of Microsoft AI, said the company’s diagnostic AI, called MAI-DxO, successfully solved 85.5% of 304 complex real-world cases published in the New England Journal of Medicine. By comparison, a group of experienced physicians reached correct diagnoses in only 20% of the same cases.

“We’re taking a big step towards medical superintelligence,” Suleyman wrote on X. “AI models have aced multiple choice medical exams – but real patients don’t come with ABC answer options. Now MAI-DxO can solve some of the world’s toughest open-ended cases with higher accuracy and lower costs.”

The technology, described Monday in a Microsoft AI blog post, was designed to emulate a virtual panel of physicians with different diagnostic approaches, collaborating to reach conclusions while also factoring in testing costs to avoid unnecessary procedures.

While generative AI systems such as GPT have previously achieved near-perfect scores on medical licensing exams, Microsoft said it set a tougher benchmark by testing MAI-DxO on sequential diagnostic challenges drawn from the NEJM’s Case Records of the Massachusetts General Hospital, detailed narrative cases often requiring multiple specialists.

This is “just the first step on a long, exciting journey,” Suleyman said, adding that the company plans further testing with healthcare partners before pursuing broader use. The system is not yet approved for clinical deployment.

According to Microsoft, MAI-DxO not only produced more accurate results than both individual AI models and physicians, but also did so with lower overall testing costs. The company said it tested multiple frontier models, including GPT, Claude, Gemini, and others, and that MAI-DxO boosted performance across each.

Will it replace physicians?

While Microsoft executives say AI will not replace physicians, they argue it could eventually augment doctors’ decision making, automate routine diagnostic work, and reduce unnecessary healthcare costs. In the United States, health spending is nearing 20% of GDP, with up to a quarter considered waste, according to Microsoft’s blog post.

The company stressed that clinical validation, safety testing, and regulatory review will be required before any real-world use of MAI-DxO. It sees the system as an early research advance towards building medical AI tools that are trustworthy, effective, and cost-efficient.

“We strongly believe that the future of healthcare will be shaped by augmenting human expertise and empathy with the power of machine intelligence,” Microsoft said.