Google AMIE transitions from simulation to real-world clinical practice with 100-patient study

Google AMIE Completes First Real-World Clinical Trial

Google AMIE, the Articulate Medical Intelligence Explorer, has moved beyond simulated environments into actual clinical practice for the first time1

1

. In collaboration with Beth Israel Deaconess Medical Center in Boston, researchers conducted a prospective study with 100 adult patients who interacted with the AI agent before their scheduled appointments1

1

. The study, published in The Lancet, marks a significant step in evaluating AI in healthcare within real-world workflows rather than controlled simulations1

1

.

The AI agent for patient interviews conducted conversations up to 5 days before scheduled primary care appointments, gathering medical history and generating concise summaries for physicians1

1

. Of the 100 participants, 98 attended their appointments, providing researchers with substantial data on how AI for clinical reasoning performs in practice1

1

. A trained physician supervised every interaction to ensure patient safety, though intervention was never required2

2

.

Clinical Evaluators Rate AI-Generated Medical Summaries Favorably

The results demonstrate that supervised use of this large language model enabled safe and meaningful clinical conversations within existing healthcare workflows1

1

. Clinical evaluators rated AMIE's conversations favorably in 87%-100% of cases, while patients rated them positively in 48%-96% of interactions1

1

. Notably, patients' attitudes toward AI in healthcare improved after interacting with AMIE and remained favorable following their medical consultation1

1

.

Primary care physicians completed follow-up surveys for 60 of the 98 cases1

1

. In 44 instances, doctors reviewed AMIE's transcript before seeing the patient1

1

. Physicians reported the transcript helped them prepare for the visit in 33 cases and might have influenced their clinical approach in 25 cases1

1

. These findings suggest the AI agent provides tangible value in supporting doctor-patient interactions and treatment planning.

How AMIE Enhances Collaborative Care

Peter Brodeur, MD, cardiology resident at Beth Israel Deaconess Medical Center, explained that AI in healthcare excels at giving patients time and space to be vulnerable and ask sensitive questions1

1

. From the clinician's perspective, he emphasized AMIE's ability to listen to patients, summarize information concisely, and ask if anything was missed1

1

. Roma Ruparel, senior user experience researcher at Google, noted that clinicians benefit from AMIE's summaries by spending less time extracting data and focusing more on collaborative care and shared decision-making1

1

.

Research scientist Mike Schaekermann explained that AMIE is more targeted and thorough than standard intake forms2

2

. For a patient presenting with back pain, while an intake form might ask for basic descriptions, AMIE can conduct follow-up questioning about pain type, onset timing, and exact location2

2

. This depth gives doctors better understanding before the appointment, allowing more time for diagnosis and developing treatment plans with patients2

2

.

Limitations and Future Research Directions

The study emphasized that results do not support standalone use of AMIE for diagnosis, triage, or patient management1

1

. José Miguel Cacho Herrera, digital health strategy consultant, described this as moving beyond exams and simulated patients, though still a limited evaluation in a real clinical setting with physician oversight1

1

. The publication in The Lancet subjects AMIE's results to peer review, providing initial evidence of conversational AI feasibility while acknowledging the need for larger studies with comparison groups1

1

.

Research scientist Anil Palepu indicated the team is working on a follow-up study across the entire United States with more patients and controlled comparisons2

2

. Google is also expanding AMIE's capabilities beyond text chat to include image sharing and real-time audio and video conversations2

2

. This evolution will allow researchers to test how the system performs when patients talk to it and show what they're experiencing visually2

2

. The evidence so far demonstrates benefits for patients, doctors, and potentially the healthcare system as a whole, though rigorous evaluation continues to build trust with the medical community2

2

.