Fabric

Conversational artificial intelligence is one component of an assembled platform, and the company is explicit that it is paired with physician authored clinical logic and customizable decision trees under the name Hybrid AI. Decision trees are rules, not models, and a customizable one is configured by the customer.

The company's own history points the same way. The conversational capability arrived by acquiring GYANT in 2024, the asynchronous care engine arrived by acquiring Zipnosis, and the clinical network arrived by acquiring a physician group's virtual service line. What is being sold is consolidation of a fragmented technology estate, which is a real and valuable proposition, and the artificial intelligence is a feature of it rather than the thing itself.

Routing is automatic, and nothing published describes a person checking it. Fabric's triage page says the system identifies acuity levels and automatically routes patients to appropriate points of care, sending those who need immediate care to the best in-person location, through 24-hour AI chat and a self-service symptom checker.

The clinical logic is physician-built and customizable by the customer, which constrains what the conversational model can do, and clinicians deliver the care that follows, including Fabric's own 50-state network in virtual visits. Post-visit workflows escalate at-risk patients.

The company does not say whether any routing decision is reviewed before a patient acts on it, which parts of a conversation the model drives and which follow the authored logic, or what happens when a patient describes something the logic does not cover.

The architecture is described in general terms, and nothing in it is identified. Fabric calls its approach Hybrid AI: advanced conversational AI combined with physician-built clinical logic and customizable decision trees, with peer-reviewed protocols that customers can adapt. The conversational layer came from its 2024 acquisition of GYANT and the asynchronous care engine from Zipnosis.

No model, provider, version, training description or triage performance figure is published: no accuracy for acuity assignment, no undertriage or overtriage rate, and no account of how protocols were validated beyond who wrote them.

Nothing establishes who else sits between a patient's conversation and an answer. Fabric names no conversational model, model provider, hosting provider or subprocessor on its home, triage, security or privacy pages, and publishes no subprocessor list or the retention and training terms that would apply to them. Its Hybrid AI is described only as conversational AI combined with physician-built logic, so a buyer cannot tell whether a third-party model receives patient symptoms. The parties handling patient content also include Fabric's own clinical network, whose clinicians see intake data in virtual care.

Named customers and vendor percentages, without the method behind them. Fabric's triage page reports more than 10 percent of patients using self-service chat, 35 percent shorter contact center waits, 30 percent lower contact center volume and a 10 percent reduction in readmissions, without naming the customers, periods or baselines.

July 2025 case studies cover Luminis Health's emergency department throughput, Hackensack Meridian Health's 89 percent screening completion with the virtual assistant, and MLK Community Healthcare's conversational AI; Intermountain Health, OSF HealthCare and Rush University System for Health are named partners. Nothing published evaluates triage accuracy, the result that matters most for a product that routes patients by acuity.

General assurances are published, and the questions a triage and chat product raises are not answered. Fabric's privacy policy, updated September 2025, covers its patient-facing solution and says it may use information to improve its services and keeps personal information for as long as necessary for the purpose collected. The security page describes logical and physical access restrictions, change management and operations monitoring.

Nothing says whether patient conversations or symptom data train or tune models, whether data is de-identified, or how long chat transcripts and virtual visit records are kept. Fabric is also a care provider through its clinical network and the Evo service, so in some pathways it holds patient records as the treating organization rather than on a health system's behalf. Safety engineering, meaning guardrails, hallucination mitigation and how a safety event is handled, is described only through physician-built logic.

Business associate status is stated in a privacy document that reaches the service, and the agreement itself is not published. Fabric's privacy policy, updated September 2025, covers both its website and its cloud-based solution for patients, and says that where it collects protected health information on behalf of a healthcare organization it acts as that organization's business associate, with the organization's notice of privacy practices and Fabric's agreements governing the data.

The security page states HIPAA compliance. No business associate agreement, its terms or subcontractor obligations are published. The dual role is not addressed: in the Evo virtual care service and its clinical network Fabric delivers care itself, which makes it a provider rather than a business associate for those patients.

Fabric names a SOC 2 Type 2 attestation as its own, with nothing retrievable behind it. Its security and compliance page says the platform is fully SOC 2 Type 2 certified and HIPAA compliant, describes access controls, change management and operations monitoring, and commits to WCAG 2.1 accessibility across its interfaces. An earlier integrations page also named HITRUST, which the security page does not. No report period, auditor, request route, trust center or penetration testing disclosure is published.

Fabric markets automated triage and states no regulatory position. Its triage page says the system identifies patient disposition and acuity level and automatically routes patients to appropriate points of care, sending those who need immediate care to the best in-person location, and describes decision support as speeding up diagnosis.

Sorting patients by urgency and directing them to emergency, urgent, virtual or self care is triage done by the software, and the routing reaches the patient without a described clinician review. The clinical logic is physician-built and peer reviewed by Fabric, and customers can customize it, but it is Fabric's own protocol set rather than a named, published triage protocol. A named, published protocol with a clinician's disposition behind each routing, or a stated regulatory position matching this capability, would change the reading.

Governance is described as physician authorship and peer review of the clinical logic, with no evaluation behind it. Fabric says its protocols are physician-built, peer reviewed and customizable, supported by human expertise, and that its interfaces meet WCAG 2.1 accessibility standards, which matters for a digital front door used by patients with visual, motor or cognitive impairment.

Nothing is published on subgroup performance, model monitoring or triage error rates, although the system decides where patients enter care at large scale. Triage is a well documented site of disparity, with differences in acuity assignment recorded across race, language and presentation style, and a conversational system applies whatever its logic and models carry to every patient.

Mechanisms catch errors in care that follows triage, and nothing catches an error in the routing itself. In Fabric's virtual care pathways a clinician reviews the patient's gathered symptoms before treating, and post-visit workflows escalate at-risk patients. The routing decision that sends a patient to self-care, a virtual visit, urgent care or the emergency department has no described review, and no warranty, remedy or error rate is published.

The two directions of triage error are not equally visible: overtriage costs money and is noticed, while undertriage sends someone to a lower level of care and surfaces later as a deterioration that is rarely traced back. Fabric is also the treating provider in its own clinical network, which gives it a different liability position there than as a software supplier.

Integration is described through standards, with no record system named. Fabric's triage page says it integrates to read and write patient records through the customer's existing record system, and its integrations material cites HL7, FHIR and API connections into record and other systems.

Large deployments at Intermountain, OSF HealthCare and Luminis Health imply working connections, but no record system vendor, marketplace listing or certification is named, and the write-back is not described at the level of what lands where.

A single hosted service for US use, with nothing committed about where it runs. Fabric's privacy policy says its services are intended for use only in the United States, and its security page describes logical and physical access restrictions without naming a hosting provider or region. No tenancy model, residency commitment or retention schedule for chat transcripts, recordings and virtual visit records is published. Those are different classes of record, and in the virtual care pathway Fabric's own clinicians create them, so where they rest and under whose control matters.

Cost stays out of view until a sales conversation. Fabric sells to health systems, health plans, employers and brokers through demo requests, and its offer spans a provider directory, a virtual care platform, an AI assistant, patient engagement and, through the Evo service, virtual care delivered by its own 50-state clinical network. No price, unit of charge or pricing basis is published for any of them. The purchase may be software, clinical care or both, which carry different cost structures.

Coverage is broad and named, with validation behind part of it. Fabric serves health systems and hospitals, health plans, employers, brokers and veterans, across emergency department intake, in-person and virtual care, asynchronous and synchronous visits, talk therapy and psychiatry through its clinical network, and post-visit engagement.

Case studies and partners name Luminis Health, Hackensack Meridian Health, MLK Community Healthcare, Intermountain Health, OSF HealthCare and Rush University System for Health; clinical scope centers on low-acuity and undifferentiated presentations. Services are US only. Its provider directory does administrative work, search and scheduling, guiding patients to the right care with customizable decision trees and streamlining scheduling. The symptom gathering and triage alongside it are clinical.