For years, AI leaders have touted healthcare as one of the clearest examples of how their technology will improve lives. Now, “it’s finally here,” says Mustafa Suleyman, CEO of Microsoft AI. The edge. The company claims to have developed an AI-based tool that it says can diagnose diseases up to four times more accurately than human doctors.
The results, published Thursday in one of the most prestigious journals in the world, Scienceare based on tests pitting Microsoft’s AI system against human doctors over a series of scenarios drawn from medical case reports and emergency department data. Microsoft compiled the cases into its own benchmark, SDBench, which it plans to make open source. Designed to function like a panel of medical experts, Microsoft’s Diagnostic Orchestrator uses multiple AI agents to reason about symptoms and medical history, decide which tests to order and arrive at a diagnosis, mimicking the process doctors use to investigate cases.
On cases taken from hundreds of reports published in The New England Journal of MedicineMicrosoft’s system achieved a diagnostic accuracy of around 83 percent, compared with around 36 percent for GPs using referrals and 22 percent for those working without them. The Diagnostic Orchestrator also outperformed doctors on emergency department cases drawn from a large database of patient admissions at Beth Israel Deaconess Medical Center in Boston, with an accuracy rate of 80 percent, compared to an average of 47 percent for human doctors. The AI system also tends to make its diagnoses at a lower estimated cost than its human counterparts.
The researchers say the system, which is powered by OpenAI’s GPT-5, also outperformed state-of-the-art AI models running on their own, although the peer-reviewed paper does not include newer models like GPT-6 Astra, Meta’s Muse Wonder Full, Google’s Gemini 4 Argon or Anthropic’s Opus 5.5. The Diagnostic Orchestrator was first developed when GPT-4.1 was the most advanced model available, but the researchers say subsequent testing suggests that “improvements in model capacity do not necessarily diminish the benefits” of multiple agents collaborating.
Microsoft recognizes that there is a long way to go before the technology can be widely deployed. Although the tests more closely reflect how doctors diagnose conditions — by asking questions, weighing possible explanations and ordering tests before reaching a diagnosis — the system has only been tested on previous medical cases and it remains unclear how well it would work in real-world clinical settings. “This requires clinical validation,” says Suleyman. “It’s a step forward, but it’s a major step in this journey.”
The research is part of a broader Microsoft ambition to develop what the company calls “medical superintelligence” – AI capable of outperforming the best human-led healthcare systems. Microsoft used the term “superintelligence” long before Donald Trump and his administration co-opted it to replace “AI.”
Diagnostics is only part of Microsoft’s vision for health, says Dr. Dominic King, the company’s head of AI health, who co-authored the study. In a blog post published Thursday and in an article from BMJ Digital Health and AI In a medical journal, the company describes five other areas it believes AI will need to master, including coordinating clinical care, making treatment decisions, and managing the human side of medicine.
The king tells The edge that Microsoft plans to begin testing the system in a hospital setting with Beth Israel Deaconess Medical Center to see how it performs in real-world clinical environments. Neither King nor Suleyman would commit to a firm timeline, although King says the project “will likely start in the next few months.”
Commercialization could also be on the horizon, although Microsoft has not yet said when or how the technology might be available. The company already has a strong presence in the U.S. healthcare sector with its digital tools, and healthcare is one of the most popular uses of its consumer-facing chatbot Copilot. King emphasizes the need for responsible deployment, emphasizing the high regulatory and compliance requirements in medicine that should be met.
“Next year we will see extremely impressive deployments of intelligent systems in production,” says Suleyman. “I think it’s just going to dramatically change the way care is delivered.”
Suleyman has previously spoken about his “human-centered” vision for AI and said medicine is no exception. Although he predicts the arrival of “something indistinguishable from human performance in medical superintelligence in the coming years,” he insists that human clinicians should and always will have a central role in patient care.
While he acknowledges that AI could result in more work being placed on doctors, he hopes it will take some of the “drudgery” out of medicine, allowing clinicians to spend more time with patients.
“I’m very optimistic about the future,” he says. “I think that’s why we build technology.”
Gn bussni
