Beth Israel Lahey Health

Beth Israel Lahey Health uses an internally developed AI model that predicts whether a patient is likely to miss a scheduled appointment. The tool is used by BILH's IT department to support scheduling decisions and patient outreach.

Described publicly by BILH's IT team at the 2025 AI and Machine Learning Symposium, the model predicts no-show likelihood based on appointment data. BILH has stated that the intended use of this prediction is to trigger patient reminders and address barriers to attendance rather than to overbook slots, with an explicit focus on reducing healthcare disparities. The full technical scope, input features, and deployment scale are not publicly detailed beyond this disclosure.

Beth Israel Lahey Health deployed Heidi, an ambient AI scribe tool, system-wide to all of its more than 6,000 physicians across 14 hospitals and 175 primary care practices. The rollout followed a six-month pilot with 1,000 providers that showed strong satisfaction and time savings.

Heidi listens to patient-clinician conversations in real time and automatically generates clinical notes, capturing medical terminology and producing structured documentation for the electronic health record. During the pilot, clinicians saved roughly 70 minutes per day on average, and 89% reported satisfaction with note quality. BILH announced the full system-wide expansion in April 2026 after the pilot results.

Beth Israel Lahey Health integrated an AI-powered internal chatbot called ChatPPGD that lets clinical staff search more than 3,800 care guidance documents by asking questions in plain language. The tool is used internally by attending staff and fields over 800 queries per week.

ChatPPGD was built by BILH's Innovation Lab using Microsoft Azure OpenAI Service, Azure AI Search, and Azure AI Document Intelligence. Staff type natural-language questions and receive answers drawn from policies, procedures, guidelines, and directives, with references to the source documents and medical citations included. The system has shown 98% accuracy in document responses. The Microsoft customer story describing this deployment was published in February 2025.