LEARN MORE
From EHR to AI Platforms
How Stanford Health Care Cut Data Retrieval Time by 95%
Summary
Stanford Health Care (SHC) has been a participant in the CHIME Digital Health Most Wired® (DHMW) program for several years, progressively advancing as its digital capabilities matured. In 2025, SHC achieved Level 10—the program's highest recognition—for the first time, reflecting its innovative approach to digital strategy and its commitment to building a high-performing, person-centric digital health ecosystem. This digital maturity has enabled SHC to move to the next frontier: enterprise AI adoption and advanced analytics capabilities.
While clinicians were already experimenting with generative AI, the tools lived outside the EHR, added friction, and couldn't scale safely. Rather than chasing one-off pilots, the organization took a different path: building ChatEHR, a real-time, in-Epic AI platform designed around how clinicians actually work.
Embedded directly into clinical workflows, ChatEHR allows staff to ask plain-language questions, automate routine tasks, and surface patient-specific insights without leaving the chart—all while maintaining rigorous governance, privacy, and accuracy standards. Supported by MedHELM, Stanford's
purpose-built evaluation framework for clinical AI, the platform transformed generative AI from an experiment into an enterprise capability, reducing cognitive burden, accelerating workflows, and laying the foundation for responsible, scalable AI across the health system
40-70%
Workflow Time Reduction
Reduction in time spent on targeted
clinical workflows such as SBAR review
From ~2 minutes to ~4 seconds
>95%
Data Retrieval Time Reduction
From a single automation
(Sequoia Transfers)
4+
Hours Saved Daily
The
Clinicians across Stanford Medicine were spending significant time navigating the EHR to locate, synthesize, and document patient information. While interest in generative AI was growing rapidly, most tools lived outside the EHR—forcing time-inefficient copy-paste workflows, introducing compliance risk, and limiting real-time clinical usefulness.
At the same time, Stanford Medicine faced rising demand for AI-powered automation across care delivery, quality, and operations. Point solutions and one-off pilots proved difficult to scale, and traditional reporting systems could not support real-time, in-workflow decision support. The organization needed a secure, governed way to bring AI into the EHR without disrupting care or clinician trust.
Outside-EHR
AI Tools
Forced time-inefficient copy-paste workflows, introduced compliance risk, and limited real-time clinical usefulness
Scaling
Difficulty
Point solutions and one-off pilots proved difficult to scale across the enterprise
Reporting
Limitations
Traditional reporting systems could not support real-time, in-workflow decision support
Bring AI directly into the EHR
Reduce time spent searching, reviewing, and documenting
Establish enterprise-wide AI governance and evaluation
Create a reusable platform that could grow with demand
Standardize evolution and monitoring of generative AI performance across use cases
Stanford Medicine set out to achieve the following goals:
The Solution
Stanford Medicine developed ChatEHR in-house as a secure, enterprise AI platform embedded directly within Epic. It enables clinicians and staff to query patient data in plain language, automate repeatable workflows, and surface insights – all while maintaining patient context, privacy, and governance.
Built
Into Epic
Built directly into Epic, so clinicians can use it in real time without leaving their workflow
FHIR-Powered Context
Pulls the right data at the right moment using FHIR, keeping responses tied to patient context
Multi-Model Architecture
Uses multiple AI models behind the scenes so the platform can adapt as technology, regulations, and clinical needs evolve without disrupting workflows
