Garner Health
Garner Health ranks doctors on the quality and cost of their care, then pays employees to see the best ones. Its rankings come from claims covering about 320 million patients across commercial, Medicare and Medicaid plans, scored on more than 650 clinical and financial metrics in 88 specialties.
Employers add Garner on top of the health plan they already have, with no network change. Members find a Top Provider in the app or through the Concierge team, and a health reimbursement account or a first dollar HSA covers eligible out of pocket costs at that doctor. Garner says the share of employees seeing Top Providers rises from 23% to 43% in the first year. An Aon analysis of 2020 to 2024 claims found first year medical spending 7.4% lower than in a matched control group, about $345 per member.
Health systems and provider groups buy the same data as Garner Insights, through an API, a referral interface or custom analysis; Mercy, Atlantic Health, Teladoc and Marathon Health use it.
Two AI agents arrived in May 2026. Garner Assistant answers benefits questions and finds doctors, and Garner Research Agent turns new medical literature into candidate metrics. Founded in 2019 and based in New York, Garner serves more than 700 organizations and 2.5 million members. Employers pay per employee per month, with up to 300% of fees at risk against a savings target.
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
Most of what an employer buys is a claims based ranking of doctors, an incentive account and a human Concierge team. AI sits at two points inside that. Garner Research Agent, launched in May 2026, reads new medical literature and drafts candidate metrics for judging physician performance, working alongside the clinical research team that maintains the measures. Garner Assistant, launched the same month, answers member questions about benefits and coverage, finds in network doctors with openings and can add one to a member's care team.
Data science and AI also keep the doctor directory current, and Garner puts its accuracy on phone numbers, addresses and availability at 92%. The rankings themselves are built from claims measures.
Member questions can be settled with no person involved. Garner Assistant answers benefits and coverage questions, finds in network doctors and can add a recommended doctor to a member's care team on its own. When a question is complex or beyond it, the assistant routes the member to a Concierge specialist and passes along the full conversation, so nothing has to be repeated. It does not diagnose or treat.
The research agent works the other way round: its literature findings go to the clinical research team, which owns the measures. No rule says when the assistant must hand off beyond the questions it cannot answer, and specialists do not see answers before members do.
Garner explains its technology through its data. Rankings draw on commercial, Medicare and Medicaid claims, enriched with hospital and health plan price transparency files, and score each doctor on specialty specific measures of diagnosing, prescribing, tests, procedures and patient outcomes, developed with what Garner calls a bottom up method. A provider profile shows a composite score with its component parts.
The two agents are described by their jobs, one reading literature and proposing metrics, the other answering members. Which language models run them is not named, and neither the formula that combines a score's parts nor the way a new metric or model version goes live is laid out.
The data supply gets far more attention than the model supply. Garner says its claims cover about 75% of US claims, more than 60 billion medical records from 320 million patients, combined with public price transparency files, though the claims sources themselves go unnamed.
The companies behind the language models in Garner Assistant and Garner Research Agent are not identified either, and no subprocessor list or model provider terms sit alongside the product pages. For an employer, that leaves open where a member's benefits conversation travels once it leaves the app, and whether the model provider keeps it or trains on it.
An Aon analysis, summarized by Garner in March 2026, compared Garner eligible members with a matched control group across several employers from 2020 to 2024, matched on geography, demographics and comorbidities. First year medical spending came in 7.4% lower, about $345 per member.
Garner's own program figures go further: 75% of employers lower medical trend by more than 5% in year one, the share of employees seeing Top Providers rises from 23% to 43%, and engaged employees take 2.7 fewer sick days a year. Those company figures come without a method. The Aon result covers the whole benefit program and ends before the AI agents launched, and the full report is released only on request.
Two kinds of data move through Garner. The provider rankings are built from deidentified claims, while the member app holds identified health, claims and reimbursement details for every covered employee. For that protected information Garner works as the employer's business associate, and the employer agreement governs its use, so the broader permissions the privacy policy grants over other data do not reach it.
Member conversations with Garner Assistant are the open question. How long prompts are kept, whether they stay apart from research work, and whether any model trains on them are not addressed, and no testing for wrong answers or process for reporting one is described for either agent.
Protected health information flows to Garner as a business associate of each employer. The service privacy notice explains that the employer agreement limits how that information is used and disclosed and requires safeguards, and it covers the member service itself rather than website visitors alone. Employer data arrives through an eligibility file, and members add claims and receipts themselves in the app.
The business associate agreement is not public, so its subcontractor terms, breach notice window and return or destruction terms are not visible before contracting. Health systems and provider groups buying Garner Insights sign a different contract, and nothing on the site says what their terms cover.
Garner completed a SOC 2 Type II audit in July 2023 and shares the report on request under a nondisclosure agreement. The announcement describes secure development and operating controls but gives no audit period, scope or exceptions, and no newer report has been announced.
That matters more now than it did in 2023. The app holds identified claims and reimbursement records for more than 2.5 million members, alongside the deidentified claims behind the rankings, and Garner Assistant and Garner Research Agent, both launched in 2026, may or may not sit inside the system that was audited.
Care navigation and benefits are the assistant's whole scope: finding a doctor, explaining coverage and tracking reimbursement inside the member app. Garner states plainly that it does not diagnose or treat. Garner Insights compares doctors to inform referral and network decisions, and it does not produce a diagnosis or a treatment recommendation for any patient.
No device clearance applies to either product, and Garner does not cite a specific regulatory exclusion. A doctor's score does reflect how well that doctor diagnoses and prescribes, but the score is used to pick a doctor, not to make a clinical decision about the patient in front of one.
Fairness work shows up in the doctor comparisons. The rankings leave out the most complex cases and adjust the rest for comorbidities and demographics, so a doctor who treats sicker patients is not marked down for it. How many cases the exclusions remove, and whether some patient groups end up thinly represented in a doctor's score, is not reported.
The AI agents come with no stated governance: no account of how a model is approved, how a change is reviewed, or whether the assistant's answers hold up equally across member groups and languages. The research agent's proposals pass through the clinical research team, which is the one human check described.
Employer contracts put Garner's own money behind the results. Fees are tied to a net medical trend target, and groups of 500 to 5,000 enrolled members can have up to 300% of fees guaranteed against it, larger groups up to 100%, with any surplus returned to the employer in full.
That commitment covers the program's savings as a whole. A member who gets a wrong benefits answer from the assistant or a poor doctor match can go to the Concierge team, but no remedy for an individual error is set out, and a program that meets its target can still carry mistakes in single cases.
Health systems reach Garner Insights in three ways. An API, now at version 1.11 and secured with OAuth 2.0 tokens, embeds provider recommendations in an existing care navigation or referral tool. A ready made interface lets care coordinators and referral teams filter doctors by specialty, network, location, language and appointment availability. Custom analyses cover anything else.
None of these is tied to a named EHR such as Epic or Oracle Health, and nothing writes back to the patient chart. The claims behind the rankings are input data rather than an EHR feed, and employers connect through an eligibility file.
Delivery is hosted: employers, members and provider customers reach Garner through its web and mobile apps or its API. The privacy policy places storage and hosting in the United States while allowing some processing outside the country. No private or customer hosted option is offered for the provider analytics or the member app.
Where the AI agents process a member's question, and whether one customer's data is kept apart from another's, are not stated. Rollout itself is light, since an employer sends an eligibility file and Garner's account managers handle implementation, education and open enrollment materials.
Employers pay a per employee per month fee that covers administration, broker compensation and the most Garner will pay out in incentives. The fee is tied to an agreed savings target, and unused incentive money goes back to the employer.
No rate appears on the site. Garner models the fee for each population and benefit design, and offers a free savings calculator to start the conversation. Provider analytics, API access and the referral interface are sold through a separate route, also with no listed price. Members pay nothing, since the program comes to them as an employer benefit.
US employers and the members of their health plans are the core market, and Garner works with any plan type without changing the network. Measures cover 88 specialties, from primary care through medical and surgical fields.
Named employer customers include the City of Portland, Mohawk Industries, Advance Auto Parts and Clayton Homes, and Sheltering Arms reports nearly 60% of its people signing up. Garner also sells through health plans and benefits advisors. Among care providers, Mercy, Atlantic Health, Teladoc and Marathon Health use the analytics, and Mercy and the University of Oklahoma appear as partners.
Alternatives to Garner Health
Other products that do the same kind of work as Garner Health, and what each one does.
Cedar Gate Technologies runs value based care arrangements end to end on one platform, from data management and analytics through to adjudicating the payment.
mPulse reaches health plan members at very large scale and uses models to decide which members to reach.
Upside (UpsideHOM, Inc.) is a housing stability service sold to Medicaid and Medicare Advantage health plans, health systems and employers, addressing housing instability as a driver of avoidable emergency and hospital use.
NOF1 is a healthcare intelligence platform that gives health plans, provider organizations and healthcare investors a working view of coverage policy, provider contracts and market data.
Arcadia sells the data foundation underneath value based care, and increasingly the models that run on top of it.
Transcarent is a health benefits navigation and care platform for self insured employers, health plans and labor unions, built around an AI navigation layer with care services around it.
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 employer fee; provider analytics quoted separately
|
Employers pay a per employee per month fee that covers benefit administration, broker compensation and incentive payments, priced for the covered population and benefit design. Provider analytics, API access and referral tools are sold separately, with no listed rate. | Garner acts as each employer's business associate for member health information. Terms for provider analytics customers are not public. | No separate implementation fee is listed; account managers handle setup, employee education and open enrollment materials. | Vendor Published |
Employer contracts can put fees at risk against an agreed savings target, up to 300% of fees for groups of 500 to 5,000 enrolled members and up to 100% for larger groups. The fee includes broker compensation and the maximum incentive payout, and unused incentive money goes back to the employer. Members join free as part of their employer's benefit. Neither the employer fee nor the provider analytics price is listed.