Microsoft AI Diagnostic Orchestrator Scores 83% Across 304 Cases

Microsoft’s AI Diagnostic Orchestrator reached 83% accuracy across 304 complex medical cases, a result that will be followed by a planned study at a Boston medical center.

Microsoft AI Diagnostic Orchestrator (MAI-DxO) follows a staged diagnostic process. It starts with a patient’s main complaint, selects questions and tests, then uses incoming information to choose the next step. The approach models a clinical team gathering details over time, rather than analyzing all symptoms and test results at once.

The benchmark covered 304 complex cases drawn from case proceedings in the New England Journal of Medicine. In the tested configuration, MAI-DxO used GPT-5 with no spending limit. Its accuracy was more than twice that of doctors using textbooks and online resources.

Microsoft plans a prospective study with Beth Israel Deaconess Medical Center in Boston. Patients will interact directly with MAI-DxO, and researchers will compare its performance with physicians’ independent assessments.