Loyal Health

A patient engagement platform with a conversational layer, where most of the product works without a model. Provider and location data management, directory listings on third-party sites, ratings and reviews management, review response and review invitations, a customer relationship management layer for outreach campaigns, and online scheduling are data and workflow products.

The model work sits in chat, which the company says runs on its own AI engine with natural language processing and machine learning prediction models inside dialog trees the health system builds, and in care search, described as using natural language processing to match what a consumer types to providers and services. A health system could buy most of the platform and never touch a model; chat and search are features layered on it.

Autonomy is claimed and oversight is a configuration, not a described control. The chatbot answers questions, surfaces providers and resources, books appointments and runs virtual triage from dialog trees and preconfigured responses the health system builds with drag-and-drop tools, natural language processing and machine learning prediction models. A person is available only if the customer buys live chat, which the company describes as an add-on to the chatbot.

Nothing published defines when a conversation moves to a live agent, whether a patient can ask for one, what hours live agents cover, or what the chatbot does when a patient describes an urgent symptom outside the triage flow.

The architecture is described in general terms and nothing is identified. The chat page says the chatbot runs on the company's own AI engine, draws on a knowledge library built from more than 3 million healthcare conversations, and combines drag-and-drop dialog trees, natural language processing and machine learning prediction models with preconfigured responses; care search is described as using natural language processing and custom filters over a taxonomy-enriched provider data set.

No model type, training data description, evaluation method, version or update practice is published, and the only earlier account of the engine, a custom-built neural network at the chat's 2018 launch, has not been restated.

The model layer is stated to be in-house, which makes the chain structurally short at that point; the rest of the chain is unnamed. The chat page answers 'Does Loyal use their own Artificial Intelligence (AI) engine?' with 'Yes', citing a knowledge library built from more than 3 million healthcare conversations, and the security page says hosting is on HIPAA-compliant cloud services in the United States, without naming the provider. Zoom is named as a communication tool.

Provider and location data leave the platform by design, synchronized to third-party directories and search listings that are not enumerated. No subprocessor list is published, the clinical taxonomy behind search is not attributed, and nothing states whether any third-party language or speech service sits behind the chat.

Named deployments, several documented by the customers themselves. LifePoint Health has published that all of its more than 80 hospitals use the chat and live chat for scheduling and triage, under a five-year partnership that began in 2017.

The company's customer stories cover ScionHealth (outreach campaigns to raise Medicare Annual Wellness Visits and cancer screening), Virtua Health (chatbot and live chat for self-help and self-scheduling without calling the access center), Baptist Health Jacksonville (conversational AI to raise MyChart sign-ups and reduce call volume) and Frederick Health, and UChicago Medicine and a Virginia hospital association affiliate are named as partners.

The Clarify Health acquisition announcement describes millions of patient searches and scheduling transactions a year, which is scale of use rather than evidence of benefit. The outcome figures, periods and baselines sit inside the individual stories. No independent evaluation is published.

General assurances are published; the questions AI raises are not answered. The security page commits to protecting the confidentiality of every piece of electronic protected health information the company touches, encryption in transit and at rest, US hosting and staff HIPAA training, all within the scope of each customer's business associate agreement.

Nothing states how long chat transcripts, search behavior or outreach engagement data are kept, whether conversations train the chat's AI engine, which the company says draws on more than 3 million healthcare conversations and 'continues to grow', or whether data is pooled across health systems to improve search and chat. The platform's purpose is patient acquisition and re-engagement, which gives those answers more weight. The acquisition by Clarify Health, an analytics business, raises whether patient-level data now moves between the two.

Business associate status is stated and supported on the company's security page. It says the company secures a business associate agreement with every customer's legal team, supports HIPAA compliance within the scope of that agreement while treating compliance as a responsibility shared with the customer, and describes how electronic protected health information is handled: encryption in transit with TLS 1.2 or above, AES-256 at rest, US hosting on HIPAA-compliant cloud services, bearer-token API security, customer-administered user access, and mandatory HIPAA training for staff.

The home page adds a commitment to follow health data regulations in core practices, processes and training. The agreement itself is not published, so its terms and whether it binds subcontractors cannot be read before contracting.

SOC 2 Type 2, named as the company's own. The security page says the company has completed SOC 2 Type 2 audits for controls covering security, availability and confidentiality, with an independent third party validating its processes, and a Trust Center at trust.loyalhealth.com is linked from the site footer. Supporting controls are described: TLS 1.2 or above in transit, AES-256 at rest, bearer-token API authentication, customer-administered user access and mandatory HIPAA training.

The audit period, the auditor and whether the report can be retrieved from the trust center without a sales conversation are not stated, and no penetration testing, vulnerability disclosure or subprocessor information appears on the security page. Because the product runs inside the health system's own web pages, its script security is part of the customer's attack surface.

Virtual triage is marketed directly to patients, and no regulatory position is stated anywhere the company publishes. The chat page offers to 'deliver care faster by virtually triage patients' and says 'Virtual triage helps patients determine what level of care they need, so those who require immediate medical attention know sooner'. LifePoint Health has described its hospitals using the chat for triage.

No triage protocol is named, nothing says a clinician decides the level of care a patient is steered to, and nothing describes what happens to a patient told their need is not urgent. The rest of the platform (provider data, listings, reviews, search, scheduling, outreach) is administrative, and the library of health risk assessments consists of questionnaires. Separately, a platform that runs on a health system's public website sits in the area of privacy regulation covering tracking technologies on provider sites.

Nothing is published on how the chat engine, care search or outreach targeting is governed or tested: no evaluation method, no accuracy figure, no results by group and no governance framework. The exposure is allocation rather than diagnosis. Care search decides which providers and services a patient sees for a described need, virtual triage tells a patient what level of care to seek, and the customer relationship layer decides who is re-engaged and when.

Search and triage both depend on how a person describes a need, which varies with language, literacy and cultural framing, and outreach optimized on conversion can favor patients already most able to book. The company does say provider order in the chatbot follows the order providers are loaded into data management rather than a ranking.

Nothing is published on what happens when the system is wrong. The failures here are routing and publishing errors: care search that sends a patient to the wrong specialty, virtual triage that tells a patient their need is less urgent than it is, a chatbot that cannot resolve a question and loses the patient, and provider data errors that the platform pushes out to third-party directories and search listings under the health system's name.

No accuracy commitment on provider data, correction timeline, warranty, remedy or service credit is described, and nothing says how chat transcripts are kept or produced if a patient later disputes what they were told, including when a conversation moves between the chatbot and a live agent.

Record systems are named for scheduling. The chat page says the platform supports Epic, athenahealth, Cerner Millennium and Meditech, surfaces the next available appointment through open scheduling APIs so patients can book from chat, and integrates with Epic's MyChart patient portal; the home page describes the customer relationship layer as integrated with the record system in real time, and a developer hub is linked from the footer.

The company joined Epic's App Orchard in 2018 to book into scheduling systems from chat. Provider data management synchronizes provider and location records outward to third-party directories and search listings. Writing is limited to appointments; nothing described writes a chat conversation or triage outcome into the chart. No current Epic Showroom listing is published.

Residency is stated; the provider and the data path are not. The security page says the company's hardware is hosted in the United States on HIPAA-compliant cloud services, with data encrypted at rest using AES-256 and in transit with TLS 1.2 or above. It is a single vendor-hosted service; the cloud provider is not named, and no tenant isolation model is described.

The product also runs as code on the health system's own website, patient portal and app, so what is collected in the browser and which third-party services load alongside it is part of the data path, and nothing addresses it. After the acquisition by Clarify Health, nothing says where the platform will run or whether data moves into the acquirer's infrastructure.

No price is published, and the shape of the offer is. The platform is sold as a modular set of solutions (provider data management, care search, scheduling, chat, a library of health risk assessments, customer relationship management, directory listings, ratings and reviews, review response and review invitations), with live chat an add-on to the chatbot, to health system buyers the company names by department: marketing and communications, physician services, patient experience, population health and operations.

The only route in is a demo request. LifePoint Health has described its arrangement as a five-year partnership. After the June 2026 acquisition by Clarify Health, nothing says whether the platform keeps being sold standalone or on what terms.

Its patient access work covers provider and location data management across a health system, directory listings on third party sites, care search, and online self scheduling through the website, chat and Google into the health system's scheduling system.

Coverage is named across organizations and channels: community hospitals, academic systems and multi-site networks, reached through the health system's website, patient portal, mobile app and Google, with a library of more than two dozen health risk assessments and outreach campaigns for population health and preventive care.

Validation sits behind part of it, with LifePoint Health's 80-plus hospitals, Virtua Health, Baptist Health Jacksonville, ScionHealth, Frederick Health and UChicago Medicine named. Clinical depth is limited to routing and the chat's virtual triage; no clinical specialty coverage is described.