Latent Health
Latent Health sells pharmacy intelligence software that helps health systems get patients onto specialty, infusion and retail medications faster. Its Clinical Reasoning Engine reads the chart, medication data, claims and payer criteria, pulls the evidence a prior authorization needs, works out whether a drug falls under the medical or pharmacy benefit, and drafts appeals with citations when a payer denies coverage. Pharmacists review each finding beside its source, approve the case, send it back or ask the prescriber for more. The platform runs in four stages, identify, access, treat and engage, covering eligibility audits, tracking where prescriptions are filled and, in beta, finding care gaps and checking in with patients on therapy.
A SMART on FHIR app reads records in Epic and Cerner, and a CoverMyMeds connection exchanges electronic authorizations with payers. Ochsner Health co designed the workflow over an 18 month partnership and cut appeal drafting from 30 minutes to seven, Yale New Haven Health doubled the authorizations each person handles a month, UNC Health absorbed 45% volume growth without adding staff, and St. Luke's avoids $300,000 a year in labor cost. A quality improvement study at Ohio State Wexner found approval rates of 81.9% with Latent against 80.2% without.
Rishabh Jain and Sriram Somasundaram founded Latent in San Francisco in 2022. In March 2026 it raised an $80 million Series A co led by Spark Capital and Transformation Capital, with McKesson Ventures, General Catalyst and Y Combinator. It reports more than 60 health systems, including Mount Sinai, Cleveland Clinic, Mayo Clinic and Vanderbilt, and more than five million patient medication journeys a year, sold by enterprise quote.
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
The Clinical Reasoning Engine does the knowledge work a pharmacy team would otherwise do by hand. It reads unstructured notes, labs and medication histories, interprets drug criteria from payer policies and labels, extracts the evidence that satisfies them and drafts the authorization or the appeal, and the same engine drives eligibility audits and the newer care gap and follow up workflows.
Structured transactions still carry the request to the payer through CoverMyMeds, and pharmacists decide where professional judgment is needed. Without the learned reading of narrative records and policy language, the product would be a forms router.
A pharmacist reviews each prepared case, approves it, sends it back or asks the prescriber for missing information, with findings shown beside their chart sources and the sign off recorded in the case history. The materials also describe automatic form completion and submission, and the payer still adjudicates every request.
When a health system must approve a request before it is sent, and what supervision applies to automated eligibility checks or patient outreach, are not set out as uniform rules. Clinical staff remain responsible for the decisions and escalations assigned to them.
Medical language models retrieve the most relevant narrative chart information and match it to coverage requirements, working from medications, notes, laboratory data, payer policies and drug labels. Each finding links to its source, and the case keeps an audit trail, which shows a pharmacist what the engine compared and why it reached a conclusion. The company describes the engine as reasoning through patient data, interpreting drug criteria, extracting evidence and orchestrating work across hospitals, payers and pharmacies.
The architecture, model releases, training data and any reproducible benchmark for retrieval, policy matching or the newer workflows are not published.
Latent builds its own clinical reasoning system on medical language models. The trust center names Vanta for security monitoring, CoverMyMeds is the electronic authorization partner, and McKesson Ventures, whose parent owns CoverMyMeds, invested in the March 2026 round. The SMART on FHIR app and the CoverMyMeds link move data between the record and payers, a different role from the models that read it.
None of those relationships identifies who provides model inference. No supplier and version list covers chart extraction, case generation or appeal drafting, whose infrastructure hosts each model is not stated, and nothing shows which upstream providers can receive the clinical inputs behind a case.
Ohio State University Wexner Medical Center presented a quality improvement comparison at the National Association of Specialty Pharmacy meeting in 2025. Using authorization records from CoverMyMeds and Latent, it found average approval rates of 81.9% with Latent and 80.2% in the comparison group, an observational result from one center.
Customer case studies add operational figures: Ochsner's appeal drafting fell from 30 minutes to seven and its reviews take four to five minutes, Yale New Haven doubled authorizations per person a month, UNC Health and Mount Sinai absorbed 45% and 15% to 20% volume growth without new staff, and St. Luke's avoids $300,000 a year in labor. None of it covers the beta workflows.
Prescription and clinical information comes through an authenticated SMART on FHIR connection to the health system's record, each authorization shows what it used, and pharmacists can open the source material. The security FAQ describes encryption in transit and at rest and monitoring across production and development, and fuller security documents are available on request from the trust center.
The public privacy policy covers the website. It does not say how long production chart data, generated drafts or reviewer feedback are kept, and does not give a health system a way to exclude its data from model training or set deletion.
Latent describes HIPAA compliant processing of health system data and lists HIPAA beside its SOC 2 report in the trust center. The website terms apply to the public site and say interactions needing HIPAA protection should not happen there, while the clinical product retrieves and uses protected records inside each health system's workflow.
The clinical business associate agreement is not published, no commercial tier is shown to include it, and nothing says whether processing suppliers sign the same terms. A security claim for the platform does not by itself set out who carries which contractual duty.
The public trust center lists a SOC 2 Type II report, an application penetration test and an external network penetration test, each available on request, along with encryption at rest and in transit and continuous monitoring through Vanta. Those are the right document types for a health system security review.
The summaries do not show the auditor, period or covered production environment, so the clinical scope of the report has to be confirmed by requesting it. Connections to each health system and to outside authorization networks may fall under separate controls.
The main workflow uses coded and narrative records to prepare medication coverage requests and appeals, with the pharmacy team and the payer keeping their own decisions and the system citing chart evidence for review.
The newer stages reach further. Identify finds patients who should start a therapy and Engage follows them during treatment, both in beta, and those uses can produce findings about individual patients rather than paperwork. No FDA authorization or reasoned exclusion analysis is given for them. The electronic prior authorization work falls under health information exchange and coverage rules, with no device determination published for the engine.
Every case keeps a record linking each output to its chart source and the payer criteria applied, and a pharmacist can return a case for revision or ask the prescriber for more information. Security monitoring and review are part of the operating process, and the founder has spoken publicly about the inconsistency of two people reaching different conclusions on the same case, the problem the engine is meant to reduce.
Ohio State's study looked at approval rates for one organization overall. No results are given by payer, race or ethnicity, language or another patient group, and how often each model is reevaluated after a release is not stated.
Pharmacy staff review the record and can return a case or query the prescriber before anything goes out, and the audit trail shows why a request was drafted or submitted as it was. An appeal drafted after a payer denial is part of the pharmacy's own remedy against the payer.
The published website terms limit liability for the site only, and the enterprise service terms are not public. No accuracy warranty, contractual correction deadline or reimbursement for a coverage delay caused by a software error is offered in anything a buyer can read before negotiating.
The SMART on FHIR app reads notes, medications, labs and demographics from the record, with Epic and Cerner named and Vanderbilt describing use with Epic for medication access. The CoverMyMeds pathway carries electronic requests and payer communication, and a University of California presentation describes those connections as live. Customers range from Mount Sinai and Cleveland Clinic to Cincinnati Children's and Keck Medicine of USC.
Chart reading and a working route to the payer network are both established. No named customer shows the final authorization status or appeal outcome written back into the chart, which is different from launching the app from inside it.
The service is managed and connects to each health system's record through a SMART on FHIR app, with separate development and production environments and encryption in transit and at rest. All named deployments are at US health systems.
No cloud region is given for chart data, no location for backups and no geography for the model infrastructure, and there is no customer hosted deployment or choice of storage region. Where the hospital's own EHR sits does not settle where the data is processed once the service has retrieved it.
Health system pharmacy and medication access leaders buy through an enterprise inquiry, and the public cases include multiyear agreements and rollouts across specialty, infusion and retail channels. One platform groups prior authorization, appeals, claims and pharmacy operations, and whitepapers on pharmacy centralization and infusion time to chair support the sale.
No entry fee, price per request, price per site or volume tier is published, and setup and support fees are not stated. The care gap and patient engagement features are labeled beta and carry no commercial terms of their own.
Specialty medication access, infusion prior authorization and retail pharmacy in health systems are the settings. Ochsner runs it across pharmacy and medical benefit workflows, including appeals and eligibility audits, Vanderbilt chose it for specialty and infusion, and St. Luke's, MetroHealth, UNC Health, Yale New Haven and Mount Sinai report results in pharmacy operations. Pediatric systems such as Cincinnati Children's and Akron Children's are among the customers, and the company says it works with half of the top 20 US health systems.
Results are not independently measured for every therapeutic area, and the beta care gap and follow up services have no reported outcomes.
Alternatives to Latent Health
Other products that do the same kind of work as Latent Health, and what each one does.
Prior authorization, appeals and grievances platform for health plans and pharmacy benefit managers, across medical and pharmacy benefits.
Develop Health automates medication benefits verification and prior authorization for prescribers, pharmacies and other access teams.
Forus, formerly Tandem (renamed in May 2026 with no change in ownership), is a medication access platform that automates the path from prescription to therapy at the point of prescribing.
Mandolin automates the administrative work required to start and sustain specialty drug treatment. Its agents read referrals and clinical records, investigate insurance benefits, estimate patient costs, assemble prior authorization submissions and follow claims through payer portals and calls.
Doceree’s co-pay.com service helps health systems connect a written prescription to savings, prior authorization and program enrollment.
HendrenAI builds voice and workflow agents for pharmacies, health systems and pharmaceutical manufacturers.
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
|
The company sells to health systems through an enterprise inquiry. No public rate is published for an individual authorization, site or connected system. | The trust center describes HIPAA compliant processing. No business associate agreement, tier or fee is published. | Implementation depends on the health record, pharmacy systems and payer workflow, and setup or integration charges are set in the enterprise agreement. | Vendor Published |
The offer covers a pharmacy platform with prior authorization, appeals and access workflows. Public materials do not give separate prices by module or volume.