On October 22, 2026, Microsoft published a study in the journal Science that is being discussed in every hospital boardroom this week: its AI diagnostic system, MAI-DxO, outperformed experienced physicians on some of medicine's hardest cases, and did it while ordering fewer tests.
The headlines say "AI beats doctors." The reality is more interesting and more useful, especially for Indian healthcare. Here is what the study actually found, what it does not prove, and why it matters far beyond Microsoft's press release.
What the study found
Microsoft built a new way to test diagnostic skill called SDBench. Instead of multiple-choice exams, it turns real cases from The New England Journal of Medicine and a US emergency department dataset into interactive simulations: the system starts with limited information and must decide what to ask, which tests to order, and when to commit to a diagnosis.
The results, evaluated on 304 complex NEJM cases:
- MAI-DxO reached 83% diagnostic accuracy. Physicians working without reference materials scored 22%; with textbooks and internet access, 36%.
- On more typical emergency-department cases, MAI-DxO scored 80% versus 47% for physicians.
- It got there cheaper. One configuration held 81% accuracy at less than half the cost of comparison systems, by choosing more cost-effective tests.
MAI-DxO works less like a single chatbot and more like a panel of doctors in a meeting: multiple AI agents propose diagnoses, challenge each other, order investigations, and update their view as results arrive. Microsoft says the method is not tied to one model - accuracy improved further when they swapped in newer models.
The honest caveats
Three things before anyone declares doctors obsolete:
- A benchmark is not a clinic. Microsoft itself says the real test is performance in live clinical environments. Its first prospective, real-patient study is only now starting with Beth Israel Deaconess Medical Center in Boston.
- Physicians in the study worked under artificial constraints. They could not consult colleagues, which is exactly what real doctors do on hard cases.
- Commentators note much of the gain may come from the underlying AI models rather than Microsoft's orchestration layer alone.
Even with those caveats, the direction is clear: AI is getting genuinely good at structured diagnostic reasoning, and it is getting cheaper at it.
Why this matters more in India than in America
India's healthcare challenge has never been a shortage of brilliant doctors. It is that they are concentrated in the wrong places. Specialist care clusters in metros and large private hospitals, while smaller towns and rural districts depend heavily on generalists and overworked government facilities. Patients routinely travel hundreds of kilometres, or spend heavily out of pocket, chasing a correct diagnosis.
A system that can reason through a complex case step by step, and do it while ordering fewer expensive tests, speaks directly to three Indian problems:
- Access: an AI-assisted second opinion does not require the patient to reach a metro hospital first.
- Cost: unnecessary tests and repeat consultations are a real burden when most healthcare spending is out of pocket.
- Consistency: decision support can help a busy generalist catch the unusual case that would otherwise be missed on a crowded OPD day.
India's telemedicine habit, built during the pandemic, makes this transition easier than it would have been five years ago. Patients are already comfortable consulting a doctor on a screen. A consultation where the doctor is quietly supported by an AI that has reviewed the history, suggested the right tests, and flagged the rare possibility is a natural next step, not a leap.
None of this replaces the doctor. Indian patients will, rightly, want a qualified physician making the final call. But the clinics and hospitals that adopt AI-assisted workflows early will offer something their competitors cannot: faster, more consistent diagnosis at lower cost.
What healthcare providers should do now
You do not need a superintelligence strategy. You need the basics that make tools like this usable later:
- Digitise properly. Structured patient records, not paper files and WhatsApp photos, are the foundation every clinical AI tool assumes.
- Modernise your practice software. Appointment, billing, and case-history systems that talk to each other are a prerequisite for any AI assistance.
- Watch the regulation. Clinical AI in India will be shaped by medical-council and data-protection rules; adopt tools that keep a doctor accountable for decisions.
At VAIR IT Technologies, we build software for healthcare providers, from practice-management systems to patient-facing websites and portals, for clinics across Noida and Delhi NCR. If you are thinking about digitising your practice before the AI wave arrives in earnest, get in touch for a free consultation.
