“We built ATLAS to think and most importantly reason like a seasoned doctor, not a search engine,” said company co-founder Dr. Aaron Cohen-Gadol. “This study shows AI can be both smarter and more efficient—good news for clinicians and patients alike.”
Fewer questions, faster answers
In real clinical practice, doctors uncover a diagnosis step-by-step by asking targeted questions and ordering only the tests that matter. ATLAS mirrored that disciplined approach. All previous models have been tested on multiple choice exam questions that do not simulate real-world clinical scenarios or patients’ journey.
That means ATLAS reaches the right answer sooner and with roughly 20 % fewer tests. At an average $150 per test, that’s about $500 saved per patient work-up—without sacrificing accuracy.
Why ATLAS did better
ATLAS uses advanced clinical reasoning and is trained on the latest peer-reviewed publications and guidelines and can instantly “consult” subspecialty modules (for example, cardiology or neurology) when a patient’s case demands deeper expertise. The system’s design forces it to justify each question or test it orders, encouraging guideline-concordant reasoning instead of the “guess-and-check” practice of medicine.
What it means for patients and clinicians
- Safer care: More accurate first-pass diagnoses reduce delays and mistreatments.
- Lower costs: Fewer low-yield labs and scans.
- Less clinician burnout: Doctors spend less time wrestling with AI to get a usable answer.
