Microsoft researchers reported in Science that a new “orchestrator” AI—designed to ask questions, order tests, and refine hypotheses like a clinician—was more than twice as accurate as general physicians on over 300 complex diagnostic cases and did so at lower cost. The system coordinates a virtual panel of specialist AIs to work step-by-step rather than generating a diagnosis all at once, and it outperformed doctors on emergency-room scenarios drawn from a major admissions database. Microsoft plans a real-world trial with Beth Israel Deaconess Medical Center to triage ER patients, a key step toward clinical validation. Wider adoption will hinge on prospective results, oversight and reimbursement, as regulators and hospitals weigh safety and workflow integration.































