Patient Voice Agents
C

CipherHealth

CipherHealth sells patient communication and care coordination software to hospitals and health systems, built around three product lines. CipherOutreach runs appointment reminders, preparation before care and follow up after discharge or an emergency visit by voice and text, and its Conversational AI Agents let patients answer in their own words. The agents sort each reply into categories the clinical team approved, route high risk concerns to clinicians and show staff the verbatim answer beside a summary, and they remember earlier answers so a patient is not asked the same question twice.

CipherRounds structures patient, staff and location rounding for nurses and unit leaders. AI Summaries brief staff before a visit, and Ambient Listening drafts rounding notes from bedside conversation, keeping structured summaries rather than recordings. AI Infrastructure is the shared layer underneath, connecting data feeds once, logging every write back to the chart and letting each health system decide which actions need human approval.

The platform exchanges data both ways with Epic, Oracle Health and MEDITECH and reaches more than 85 record systems through HL7, FHIR and ADT feeds. It draws on data from more than 500 hospitals and handles more than 40 million calls and texts a year. Customers include Northwell Health, Intermountain Health, Advocate Health, Banner Health, HCA Healthcare, Norton Healthcare and UMass Memorial Health. Intermountain reports nearly $15 million in savings from post discharge calls, Lenox Hill Hospital reached the 99th percentile on key HCAHPS measures after redesigning its rounding, and a UCSF study followed 18,076 patients through scripted discharge calls.

Founded in New York in 2009, CipherHealth holds HITRUST, SOC 2 Type II and TX-RAMP Level 2 certifications. Cofounder Zach Silverzweig returned as chief executive in May 2026, and the platform is sold by enterprise quote.

AI Health Index verifiedNovember 19, 2026
Compare CipherHealth with other vendors
Founded
2009
Headquarters
New York, New York
Categories
patient-facing-voice-agents
Indexed Products
CipherOutreach, Conversational AI Agents, Automated Voice Outreach, Post-Discharge Follow-Up, CipherRounds, Ambient Listening, AI Summaries, AI Infrastructure
Buyer Segments
Health systems, Hospital care transition teams, Patient access teams, Population health programs, Nursing leadership, Patient experience teams
Assessment

Capability Axes

An AI Health Index grade measures what a buyer can verify from public sources on the date shown. It is not a rating of how good the product is. A vendor can build an excellent system and grade low on an axis because it publishes nothing an outsider can check. How grades read

AI Capability
BB on AI CentralityThe model is the engine of a core module. The platform carries other value, but this capability does not exist without it.
Vendor Published

Learned language processing sits at the center of the newer work. The Conversational AI Agents interpret free text and spoken replies, sort symptoms and barriers into categories the clinical team approved, summarize each exchange for staff and decide which replies are high risk enough to send to a clinician. Ambient Listening in CipherRounds turns bedside conversation into drafted rounding notes, and predictive models flag patients likely to be readmitted, miss an appointment or disengage.

A large share of the work runs without that step. Scripted calls, keypad answers and reminders written elsewhere are delivered without interpretation, and the clinical team chooses which mode each patient group receives.

DD on Autonomy and Oversight ModelNo oversight structure is published. An absolute claim that the system does not err grades here too, because a buyer who believes it will not build the review step that would catch a failure.
Vendor Published

Clinical leadership approves scripts, prompts and escalation paths, and clinicians work from a queue showing the verbatim reply and its summary. Inside those protocols the agents act on their own: replies are triaged, high risk answers go to clinicians automatically and alerts fire without a person checking each interpretation first. The company describes the agents running approved scripts the same way for every patient, every time.

Gateway lets each health system choose which write backs to the chart need human approval, and some conversations hand off to live staff. Approval applies to the protocol rather than to each reading of a patient's words, so staff review a routed case after the agent has acted.

BB on Model and Technology TransparencyThe approach or the suppliers are named without the version and update discipline behind them. Naming a supplier is the entry to this band both here and on Model Supply Chain Disclosure, which ask different questions of the same disclosure: who receives the data, and what produces the output.
Vendor Published

The design is set out at the level of inputs and outputs. Agents interpret free text or spoken replies, classify them into clinical outreach categories, extract issues and summarize them for staff, and each result maps to the team's protocols with the original response alongside. AI Infrastructure adds a context layer that builds a clinical view of each patient, plus a conversational data tool that answers plain language questions about outreach and rounding results. Since 2023 the data science team has built, trained and deployed models on Google Cloud's Vertex AI.

Architectures, model versions and training methods for interpreting patient meaning are not given, and no benchmark compares agent readings with a clinician's across channels.

CC on Model Supply Chain DisclosureThe architecture is described and no model provider is named. Naming a hosting provider alone does not lift a record out of this band. Record the host in the note, because it matters for residency and breach scope, and grade on the model layer, which is the question this axis is named for.
Vendor Published

CipherHealth builds the outreach platform and agent controls, and in October 2023 it chose Google Cloud's Vertex AI to build, fine tune and manage its machine learning models, alongside natural language processing already used for sentiment analysis. Google Cloud also hosts the Evolve platform in the 2022 HITRUST assessment, which matters for residency and breach scope.

Those arrangements predate the generative agents, Ambient Listening and AI Summaries. No current supplier list names the language, speech recognition or voice synthesis models behind them, gives their versions, or shows which outside parties receive the patient replies and bedside audio they process.

BB on Clinical and Operational EvidenceNamed deployments with dated outcome figures and enough method to test them, or published research short of independent validation.
Vendor Published

A UCSF retrospective study followed 18,076 patients through scripted discharge calls on CipherHealth Voice and found strong engagement across age groups, with reach and follow up needs varying by age. Intermountain Health called 6,400 patients after discharge, reached 84% on the clinical question, found 43% needed an intervention and reports nearly $15 million in savings. Lenox Hill Hospital reached the 99th percentile in key HCAHPS domains after redesigning its rounding, and Norton Healthcare reports a 42% reduction in burnout.

Those figures cover scripted outreach and rounding. The Intermountain savings come without a period or calculation, and no study measures how often the agents read symptoms correctly or choose the right escalation.

BB on AI Safety and PHI StewardshipCategorical commitments are published, such as no training on customer data, without the retention schedule behind them.
Vendor Published

Data is encrypted with AES 256 at rest and TLS 1.2 or higher in transit, behind role based access, staff training, continuous vulnerability scanning and intrusion detection. Agents keep what the patient said beside the generated summary and route issues only through the team's approved workflow. Ambient Listening encrypts audio end to end and keeps deidentified structured summaries rather than recordings.

The company says its AI draws on years of patient engagement, rounding and issue resolution data. A health system has no stated way to keep its transcripts out of model improvement, and how long outreach audio, transcripts and summaries are kept, and how deletion works, are not set out for any product.

Regulatory and Compliance
BB on HIPAA and BAA PostureBusiness associate status is stated and supported by a substantive privacy document, with the agreement or its scope not fully published. For a vendor outside the United States, an equivalent regime documented to this depth grades here. A vendor whose product does not process protected health information grades here too when it says so plainly and explains the scope, with a service privacy document behind it: stating a position a buyer can rely on is the posture this axis grades, and the band above is closed to it because there is no agreement to publish.
Vendor Published

CipherHealth identifies itself as a business associate and signs business associate agreements with all covered entity clients, with execution handled on request through the security team. Its HIPAA policy covers workforce training, role based access, encryption, incident response and help fulfilling patient rights, and the company states compliance with the Privacy Rule across all products and services. AI Infrastructure carries the same agreements for health system deployments.

The agreement text is not available before contracting, and no fee or tier attaches to it. Which telephony, speech and model suppliers handle protected health information under equivalent terms is not stated.

AA on Security Certifications and Trust CenterCertifications named with their type and version and presented as retrievable artifacts, usually through a trust portal a buyer can open without asking.
Vendor Published

The Texas Department of Information Resources registry lists the Evolve Platform under TX1481024 at Level 2 through March 25, 2027, covering CipherOutreach, CipherRounds, CipherKiosk and CipherConnect. Evolve also holds HITRUST CSF certification validated against more than 400 controls, and a SOC 2 Type II report covers security, availability and confidentiality, with audits each year. Controls include AES 256 encryption at rest, TLS 1.2 or higher in transit, role based access, and continuous vulnerability scanning and intrusion detection with real time alerting.

The SOC 2 period, the HITRUST assessment type and date, and penetration test results come by request from the security team. The registry entry confirms current certification for Evolve but not whether upstream model providers sit inside that scope.

DD on FDA and Regulatory StatusThe positioning is unclear or unsupported: clinical claims with no regulatory position stated, or language implying a status the databases do not carry. An open enforcement action that names the product also grades here, whatever the positioning: an FDA warning letter with no close out letter posted, or a consent decree in force. The note says which of the two applies, and the grade returns to the positioning bands when FDA closes the action.
Vendor Published

Scheduling, reminders, rounding and most discharge outreach serve communication and care coordination. The Conversational AI Agents go further: they interpret symptoms, identify concerns and trigger clinical escalation, and a product demonstration shows the agent offering a patient a clinical interpretation before forwarding the issue to the care team. Predictive models also flag individual patients at risk of readmission.

Clinician approved protocols and human escalation surround that work, but no device classification, FDA clearance or reasoned exclusion covers learned symptom interpretation or patient facing clinical output. The line between scripted workflow and interpretation matters here because products sold under one platform reach different depths of clinical judgment.

BB on AI Governance and Bias DisclosureA governance framework with named process behind it, such as certification to an artificial intelligence management standard, or material written for a customer own review committee to evaluate the product with. A framework covering the vendor whole portfolio counts when its scope reaches this product; what it cannot supply is the product specific evaluation the band above asks for. Published results within the one group where the performance of this product is most at risk also sit here: evaluation on the question this axis asks, short of the comparison across groups the band above requires. A medical device clearance or certification, and the quality system behind it, does not reach this band on its own; it is graded under regulatory status. A regulator reviewed plan governing how the model may change does reach it.
Vendor Published

Clinical leadership approves scripts, prompts and escalation policies before the agents run, and model cards, bias monitoring, human escalation and audit trails are described as standard for every deployment. Staff see the verbatim response beside the generated summary and can rate or refine outputs, and Tailor in AI Infrastructure lets a health system try a change on a small group before a broad rollout.

UCSF's study reports reach and follow up needs by age group for scripted voice calls, not for the agents. No model card with test methods and results by age, language or accent is available to read, and no independent fairness audit covers the open response agents or their multilingual support.

CC on AI Liability and RecourseMechanisms exist that let someone challenge an output, such as audit trails, source traceability or review before commit, with nothing standing behind the output and no route for the harmed party.
Vendor Published

Each patient response carries its escalation history, so staff can trace why a concern landed in a given queue, and Gateway logs every write back to the record. Teams can give feedback on AI outputs, and a security and privacy contact handles incidents and document requests.

The terms a buyer can read before negotiating set no remedy for a missed escalation, a misread symptom or harm from a late callback. There is no accuracy warranty for the agents and no guaranteed correction time. Who follows up when a reply raises concern is set by the health system's own workflow, which leaves the clinical risk with the care team rather than with the software.

Integration and Deployment
BB on EHR and Interoperability DepthNamed systems with read access or one directional writing, or standards support with named deployments behind it.
Vendor Published

Bidirectional connections reach Epic, Oracle Health and MEDITECH, with prebuilt and certified Epic and Cerner integrations, and Allscripts, athenahealth, NextGen, eClinicalWorks, CPSI and Homecare Homebase are among more than 85 record systems reached through HL7 v2, FHIR R4 and ADT feeds. Epic messages can start inside the chart, outreach responses and issue notes return to flowsheets, and rounding summaries sync into record fields. One unnamed live customer loaded 12,000 call responses through the flowsheet link.

UCSF uses the scripted discharge calls. No named site shows both directions of the current agent workflow in production, and what returns to the chart depends on each health system's configuration.

CC on Deployment Model and Data ResidencyA single hosted option with location implied rather than committed.
Vendor Published

Evolve is hosted software, listed as a cloud service in the Texas state registry and certified at TX-RAMP Level 2 for processing Texas state agency data. An earlier HITRUST letter places the platform on Google Cloud in US West and US Central and in an Armor data center in Texas for its two year term from September 2022, and in October 2023 the company chose Google Cloud's Vertex AI to build and run its machine learning models.

No current location commitment covers audio, transcripts, backups and model requests for the agents. There is no choice of residency region and no customer hosted option for any workflow.

Commercial
CC on Commercial TransparencyNo price is published and the posture is discoverable: a buyer can establish how the product is sold and what drives the cost before contacting the vendor. Most of the index sits here.
Vendor Published

Health systems buy through a demonstration and a clinical workflow discussion, with outreach, rounding and AI Infrastructure scoped to the protocols and record systems involved. The usual starting point is one or two measurable workflows in access or care transitions, expanding from there, and a readiness assessment on the website scores a health system's care coordination before a sales call.

No price, rate per call or text, or billing unit is given for any function. Model use, telephony, record integration and staff onboarding carry no separate stated terms, and the agents have no list price apart from the wider outreach subscription, so every element is set in the enterprise agreement.

AA on Setting and Specialty CoverageWhere the product is validated to operate is named and supported, settings and specialties both, whether the coverage is broad or deliberately narrow.
Vendor Published

Hospitals and health systems use the platform across patient access, appointment reminders, preparation before procedures, follow up after emergency and inpatient discharge, chronic disease programs, medication adherence and social needs screening, plus nurse and leader rounding on inpatient units and in emergency departments. UCSF Health uses voice outreach for discharged patients and bundled episodes including orthopedic surgery, Intermountain for post discharge calls, Lenox Hill for rounding, UMass Memorial's emergency department for Ambient Listening, and Advocate Health to extend its care transitions team.

Coverage spans acute, ambulatory, home and virtual settings. The case record does not establish equal results for each language, specialty and type of escalation in the newer agent services.

Alternatives

Alternatives to CipherHealth

Other products that do the same kind of work as CipherHealth, and what each one does.

Steer Health sells an AI workforce for patient access, clinical documentation and revenue cycle work at health systems, hospitals and medical groups, packaged as three products.

EmpowerHealth runs automated outreach for health plans, health systems, pharmacy benefit managers and digital health companies.

Get Well makes patient engagement software for hospitals, health systems, health plans and government health systems, including the Department of Veterans Affairs.

Rivvi is an AI workspace for healthcare teams: a HIPAA covered alternative to general chat assistants where staff upload rosters, payroll runs, gap lists and documents, ask questions in plain language and get tables, charts and lists back.

ActiumHealth is the healthcare patient communication product of Syllable, which acquired it in March 2024 and runs it as a division.

Clarus provides patient communication software for medical practices, health systems, and home health organizations. It handles calls during office hours and overnight, routes urgent requests to on-call providers, and holds routine messages for staff follow-up.

Commercial

Pricing

Vendor-published figures are labeled as such. Figures labeled “Estimated” are derived from third-party sources and have not been confirmed by the vendor.

Entry Price Pricing Basis BAA Tier Implementation Source
Custom enterprise quote
CipherOutreach is sold by enterprise inquiry for the selected outreach uses and integrations. The company does not publish a message, minute or patient rate. Business associate agreements are signed with all covered entity clients and executed on request through the security team. No agreement fee or tier is stated. Configuration may include EHR connections, outreach protocols, rounding setup and routing, and rollouts usually start with one or two workflows in access or care transitions. No onboarding or setup fee is stated. Vendor Published

The offer spans voice, text and conversational outreach, rounding with Ambient Listening and the AI Infrastructure layer. No separate rate is given for the AI agents, the rounding tools or the included communications volume.