Google's Articulate Medical Intelligence Explorer (AMIE) has moved beyond simulations to real-world testing in an ambulatory primary care clinic. In a prospective feasibility study conducted with Beth Israel Deaconess Medical Center, the conversational AI system conducted pre-visit history taking with 100 patients before their appointments. The research marks a crucial milestone in translating AI diagnostic capabilities into clinical practice.

Patients interacted with AMIE via secure web links before seeing their primary care physicians, with human supervisors monitoring for safety concerns. The study found zero safety stops were required across all interactions, demonstrating conversational safety in a supervised clinical setting. The AI system generated transcripts and summaries that were provided to clinicians with patient consent, mirroring how trainees operate under physician supervision.

Clinical evaluators found AMIE's differential diagnoses and management plans comparable to those of physicians in blinded assessments, though doctors scored higher on practicality and cost-effectiveness. The AI successfully identified the final diagnosis within its top seven possibilities in 90% of cases. These findings suggest supervised deployment of conversational diagnostic AI is feasible and well-received by both patients and clinicians, though the system still faces limitations in accessing electronic health records and performing physical examinations.