Patient portal AI assistant
Epic Emmie (incl. Ask Emmie)
Emmie is Epic's patient-facing AI family inside MyChart. Epic describes record-grounded questions, plain-language summaries, scheduling, billing, payments, reminders, and other actions, but no single public feature list applies everywhere. Sutter currently offers Ask Emmie to signed-in patients for questions and summaries while excluding scheduling, care-team messaging, diagnosis, and record writeback. Reid separately documents preview toggles, text scheduling, billing help, imaging-result explanation, and symptom routing for eligible primary-care patients. These functions may expand practical agency, but the health system controls access and scope. Public evidence does not identify the production model, inference host, chat retention, training or secondary use, or a patient correction or appeal route. A June 2026 study of a closely related pre-release Epic MyChart assistant reported strong usability in a selected cohort, but did not systematically review outputs for errors or unsafe recommendations and does not validate production Emmie.
Public-source research has been drafted, often with AI assistance. No comprehensive human review is recorded unless the profile provenance says so.
Summary judgment · 38% toward patient-directed
Mixed, institution-led and directly patient-facing
Patients directly initiate questions and actions that may improve understanding and access. Epic and each health system still determine eligibility, configuration, models, guardrails, data handling, and success measures. Patient design input is documented, but patient governance and control are not.
Patient agency
How this tool changes agency
Epic documents record-grounded questions, plain-language result and visit summaries, scheduling, billing, payments, and follow-up support. Sutter excludes scheduling, care-team messaging, diagnosis, and record changes. Reid documents broader functions including symptom routing for eligible primary-care patients, which raises the clinical stakes.
Patients can avoid opening the assistant. Reid documents opt-in text messaging and Preview Feature toggles for some functions. Sutter's current FAQ says Emmie is available to all signed-in patients and does not document a separate opt-out. Choice for automatically presented summaries or reminders is not publicly established.
Patient-facing signals
Who does this AI serve?
Epic and health systems procure, configure, and evaluate Emmie, often using operational measures such as fewer billing messages and staff time saved. Patients may gain faster explanations and self-service, but they do not choose the platform, its model, its enabled functions, or its governance.
Can patients tell AI is involved?
Sutter and Reid clearly name Ask Emmie as generative AI and publish patient explanations. Public sources do not establish how every automated imaging summary, visit summary, reminder, or other Emmie-generated output is labeled at the moment a patient sees it.
Can patients meaningfully choose?
Patients can avoid opening the assistant. Reid documents opt-in text messaging and Preview Feature toggles for some functions. Sutter's current FAQ says Emmie is available to all signed-in patients and does not document a separate opt-out. Choice for automatically presented summaries or reminders is not publicly established.
Can patients correct or challenge what the AI produces?
Sutter provides thumbs-up and thumbs-down feedback and tells patients to contact their provider about health concerns. No public source documents answer editing, a patient-visible correction history, appeal, transcript export, or a route that requires a disputed output to be investigated. Sutter says chats do not update the medical record.
Does it help patients understand or act?
Epic documents record-grounded questions, plain-language result and visit summaries, scheduling, billing, payments, and follow-up support. Sutter excludes scheduling, care-team messaging, diagnosis, and record changes. Reid documents broader functions including symptom routing for eligible primary-care patients, which raises the clinical stakes.
Text findings
Who is left out or burdened?
Portal, deployment, language, and evidence gaps
Access depends on the patient's health system, MyChart account, digital access, and enabled features. Sutter documents use through authorized proxy access. Reid said its May 2026 release was English-only and described 14-language messaging as future work. Disability access, low-literacy performance, broadband and device burden, and subgroup safety are not publicly evaluated. The related usability study enrolled a selected, highly educated cohort at one system.
What happens to patient data?
Covered-care context, important product details undisclosed
Sutter says chats use information already visible to the patient, are not part of the legal medical record, and cannot be seen by providers. Public sources do not identify the production model or inference host, retention period, deletion or export controls, training or secondary use, downstream processors, or how these practices vary by health system. The pre-release study's GPT-4.1 and no-training statements apply to that study, not necessarily production Emmie.
Are the clinical boundaries clear?
Clear warnings, expanding practical stakes
Sutter says Emmie provides education, not personalized advice, diagnosis, or care decisions, and warns that generative AI can be wrong. Reid documents symptom routing to self-care, appointments, urgent care, or emergency care, while Epic promotes immediate imaging summaries and chart-grounded answers. Product-specific production accuracy, calibration, escalation, and harm evidence was not found.
Who defined what good looks like?
Mostly vendor- and institution-defined
Epic reports operational gains such as fewer billing messages and staff hours saved. Reid plans to track use, navigation, and friction. Patient and advisory-council input is documented, but patient governance is not. A peer-reviewed pre-release usability study found strong usability, yet it was conducted with Epic participation, used a selected cohort, and did not systematically review outputs for errors or unsafe recommendations. Independent production safety and patient-outcome validation was not found.
Review provenance
Criteria
Same HugoScore CAIHL-derived criteria for every tool. Public criteria are in site/data/criteria.json and the fuller method is in SCORING_FRAMEWORK.md.
Reviewer
AI-assisted public-source draft refreshed in OpenAI Codex. No named human reviewer is recorded.
AI / model
OpenAI Codex / GPT-5.
Human review
No comprehensive human review has been completed or claimed for this draft profile.
Review date
2026-07-21
Limitations
No authenticated Emmie deployment, Epic UserWeb access, implementation configuration, contract or BAA review, production-model identification, transcript-retention test, patient or care-team interview, accessibility or security audit, or independent production validation. The peer-reviewed usability study evaluated a closely related pre-release Epic MyChart assistant and is not treated as a product-specific production validation. Legal sources are context, not a compliance determination.
Review method
Deep public-source review of current Epic product, newsroom, and privacy pages; current Sutter Health and Reid Health deployment pages; Reid's generative-AI policy; trade reporting and a product-design interview; a June 2026 peer-reviewed study of a closely related pre-release Epic MyChart assistant; and current HHS and California governance context. No authenticated MyChart access, Epic UserWeb documentation, implementation contract, model or data-flow inspection, transcript-retention test, patient or care-team interview, accessibility audit, security audit, or independent production accuracy, safety, equity, or outcome validation.
Draft profile · Medium draft, current official and deployer sources plus a related peer-reviewed pre-release usability study. No independent production validation