For most doctors, using an AI reference tool has meant pausing mid-visit, tabbing over to a separate app, and typing in a stripped-down version of the clinical question because the software has no idea who the patient actually is. Cedars-Sinai Health System is trying to eliminate that gap. In May 2026, the Los Angeles-based system announced it had rolled out enterprise-wide access to OpenEvidence, embedding the AI-powered clinical reference platform directly inside its Epic electronic health record.
From Generic Lookup to Patient-Aware AI
The distinction Cedars-Sinai and OpenEvidence are drawing is between AI that answers a general medical question and AI that answers a question in the specific context of the patient on the screen. Under the new integration, physicians, nurses, pharmacists, and therapists can query OpenEvidence and receive answers pulled from peer-reviewed literature that are automatically contextualized against that patient’s labs, medications, diagnoses, and history sitting in Epic. OpenEvidence describes this as “agentic clinical AI,” the first time the capability has been deployed at enterprise scale according to the company.
How It Actually Works at the Bedside
Instead of a clinician manually describing a patient’s renal function or drug list to get a relevant answer about, say, dosing adjustments or drug interactions, the system pulls that data automatically from the record. The promise is speed: fewer clicks, less context-switching, and answers that are already filtered for relevance to the person in the bed rather than a generic textbook case. Cedars-Sinai says the tool is available to essentially every type of clinician in its workforce, not just physicians, reflecting a broader trend of extending AI reference tools to nurses and pharmacists who make plenty of point-of-care decisions themselves.
Privacy Guardrails and Governance
Both organizations have been explicit that patient data pulled into OpenEvidence is used only to contextualize the immediate query and is not permanently stored by the company or used to train its underlying models. Cedars-Sinai also says it runs the integration through an internal governance structure in which a standing committee of data scientists, clinical leaders, and administrators reviews and audits AI tools before they go live and on an ongoing basis afterward — a response to broader industry criticism that health systems have moved faster on AI purchasing than on AI oversight.
Cedars-Sinai Isn’t Alone
The move follows a similar integration at Mount Sinai Health System, which wired OpenEvidence into Epic across all seven of its hospitals earlier in 2026, letting its own physicians, nurses, and pharmacists ask natural-language clinical questions grounded in the literature. The near-simultaneous rollouts at two major academic systems suggest OpenEvidence, which has quickly become one of the most widely used AI reference tools among U.S. physicians, is positioning EHR-native integration as its next competitive front, rather than remaining a standalone app clinicians have to remember to open.
The Skeptical View
Not every health IT leader is sold on embedding AI reference tools this deeply into clinical workflows. Critics point out that even literature-grounded AI can surface outdated guidance, misinterpret nuance in a patient’s chart, or create a false sense of authority simply because the answer appears inside the trusted EHR interface rather than a separate app clinicians know to double-check. Governance researchers have also warned generally that health systems are deploying AI faster than they can build the monitoring infrastructure to catch it when a model’s outputs drift or degrade, a concern that applies just as much to reference tools as to predictive algorithms.
What’s Next
Cedars-Sinai says its next step is to layer its own internally developed care pathways and clinical protocols on top of the OpenEvidence integration, so clinicians can view outside medical evidence alongside the health system’s own institutional guidance in a single interface rather than having to reconcile the two separately. If that layering works as planned, it would push the tool from a literature-lookup assistant toward something closer to a full clinical decision-support system tailored to how Cedars-Sinai itself wants care delivered — a template other large academic medical centers are likely to watch closely as they weigh their own enterprise AI deals. Both organizations have said they plan to publish usage data over the coming months, including how frequently clinicians query the system and in which specialties adoption runs highest, information other health systems are watching for as they decide whether a similar EHR-embedded deployment is worth the governance overhead required to stand one up responsibly.
