Valerie Health
Valerie Health runs an AI front office for independent physician groups, combining software agents with its own operations staff. Referral Automation gathers referrals arriving by fax, email, online portals and electronic records, identifies them and extracts patient, insurance and referring provider details. Valerie's team verifies the fields, the platform requests anything missing from the referring office, and the verified referral is written into the practice's record so it can move to scheduling. Fax and Document Automation classifies other incoming faxes, such as payer responses on authorizations, and routes them to the right team. Valerie Health Communications brings phone, text, chat and email into scheduling and follow up, booking appointments and confirming insurance by the practice's rules, and Growth Analytics shows referral patterns and conversion.
At ENT Partners, referral and fax intake cut median chart creation to 26 minutes and raised same day bookings 79% during the pilot, with referrals written into eClinicalWorks. Upperline Health expanded a Florida pilot across its network with records routed into athenahealth, and Pinnacle cut its fax spending. Customers span cardiology, podiatry, urology, dialysis, primary care, ENT, neurology, home health and behavioral health, and US ENT Partners named Valerie a preferred partner in July 2026.
Peter Shalek, who cofounded Joyable, and Nitin Joshi, a cofounder of Uber Health, lead the San Francisco company. Redpoint Ventures led its $30 million Series A in December 2025, bringing funding to $39 million. Pricing comes through a practice demo.
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
Learned models make the first interpretation of everything the service handles. Vision language models identify a referral among incoming messages and extract patient, insurer and referring provider fields, a classifier sorts other faxes by type, and on patient calls speech recognition, language models and speech synthesis interpret the request and carry out scheduling or referral follow up.
Each output becomes a structured task for verification, collection of missing information and routing into the record. The operations team checks extracted data, which makes the service a blend of software and people, but the models are the engine of both products: without them there is no classification, extraction or conversation for the staff to verify.
Every extracted referral passes human verification by the operations team before verified fields sync into the electronic record, the employees in the loop model the company describes. The software checks for required information and asks the referring provider for what is missing, fax parsing assigns each item and its next step to a member of the practice team, and automated scheduling follows rules the practice supplies. That is a described oversight structure with a verification gate.
The parts left open are the threshold at which an uncertain classification pauses, the override path after something has been written to the chart, and the set of patient requests the communications product always passes to staff.
Vision language models extract information from referral forms and other documents, a machine learning classifier sorts each fax or message by type, and the communications system pairs speech recognition and speech synthesis with language models to interpret calls and produce replies. The trust center names Anthropic as a subprocessor without assigning it to a voice or document task. That identifies the approach and one supplier. The deployed model versions, the data the classifier and extractors were trained on, and a release history showing when a model change could alter classification or transcription are not given.
The trust center's subprocessor list names Anthropic in its own category and Google Cloud Platform and Supabase as cloud providers, and the document and communication products describe vision, language and speech models. That names the model provider and the hosting layer.
The mapping is missing: nothing says which provider handles document extraction, which handles speech, whether call recordings or referral images reach a separate speech vendor, or what training and retention terms apply to patient data once it arrives at each party. The list answers who is in the chain, while the terms that govern the data at each step stay with the contract.
The ENT Partners case study reports a median of 26 minutes to create a chart and a 79% increase in same day bookings in an AI assisted referral and fax pilot that processed thousands of documents and wrote into eClinicalWorks. Upperline Health's case study describes expanding a Florida pilot network wide, with referrals reaching the record in under 30 minutes at the median, and Pinnacle's chief executive reports lower fax spending.
Across customers, the company cites new patient visits up more than 5%, a sixfold return and median referral entry as fast as 19 minutes. All are vendor reports with no measurement window, baseline or comparison against clinics outside the deployments.
The service handles referral documents, patient identifiers, insurance details and communications carrying scheduling information. The trust center describes data retention procedures and states that customer data is deleted when a customer leaves, and the general privacy notice leaves the use of protected health information to the relationship with the practice.
No day count is given for holding referral files, transcripts or call audio, no deidentification method is described, and nothing restricts the models that receive patient material from training on it. There is also no procedure for handling an unsafe reply to a patient, or a wrong extraction, as a safety incident.
The service privacy policy identifies Valerie Health as a business associate when a covered customer provides identifiable patient information, separating those records from data about website visitors and healthcare professionals' accounts, and it points patients to the provider's own privacy practices. The homepage calls the service HIPAA compliant and certified, and the trust center lists a 2025 HIPAA report.
The agreement itself is not available to read, no plan is named as carrying it, and nothing describes how it reaches the operations staff who verify referrals or the subprocessors that handle documents and calls.
The trust center lists a 2025 SOC 2 Type II and HIPAA report, a 2026 SOC 2 bridge letter and a July 2026 penetration test, with controls for account access, retention procedures and disaster recovery testing. The reports and the bridge letter open only after an access request, so the audit period, the opinion and the treatment of the communications and referral products cannot be read from the portal. The documents are named by type and year. Whether voice processing at subprocessors, or the operations team's verification work, falls inside the tested scope is not shown in the summaries.
Independent practices use the software for front office administration: processing and routing referral paperwork, collecting missing information and following practice rules for scheduling and communication. The fax product can detect a payer's response on an authorization and send it to the relevant team, without making the payer's decision or recommending a treatment, and the voice product's described tasks are booking visits, confirming insurance and following up on referrals.
The company sets out the administrative users and tasks, and the product sold matches that scope. No FDA clearance is claimed and no device exclusion is cited for records handling, scheduling rules or patient communications.
Extracted referral fields are checked by an operations team, and the communications products describe the speech and language models they use, which shows how work is handled. Model governance is absent. There is no review procedure for releasing a new model, no assessment of performance for callers with different accents, speech patterns or languages, and no breakdown of error rates by site, specialty, referral format or patient group, since the case studies aggregate by practice group. The security controls in the trust center carry no model specific evaluation or governance commitment, even as the company adds voice across nine specialties.
Extracted values are checked before they are written to the record, practices can see the returned referral information in the chart, and staff can trace which item went to the scheduling team. Those steps let a practice catch corrections in ordinary work. The public terms cover use of the website and set no accuracy commitment for the referral or voice systems, no correction route for a patient given a wrong scheduling answer, and no deadline for repairing a mismatched document. Once information reaches the practice's systems, follow up is the practice's responsibility.
ENT Partners' case study describes referral fields written into eClinicalWorks, patient records updated and a scheduling notice sent when the patient is ready to book, and Upperline Health's describes verified referrals routed into athenahealth across a multisite rollout. The service can also receive referrals from an electronic record and connects to practice management and fax systems. The named deployments establish writing into two major EHRs. Neither documents a full read and write interface, separates the fields read from the record from those written to it, or points to a marketplace listing for the workflow.
The service is hosted and connects to the channels and record systems a practice already uses. The trust center lists Google Cloud Platform and Supabase as cloud providers and describes access controls for production infrastructure, which names the hosts without giving a region or tenant design. Where referral documents and call records are stored, where each model runs inference and whether any processing crosses a border are not stated. There is no option to run the models inside a practice's own infrastructure, and the embedded operations team adds people to the data path whose location is not given either.
Sales go through a demo request that asks about the practice and its record system. The service covers referral, fax, document verification, scheduling and communications work, and the company frames its value in customer results, a sixfold return, more than 5% growth in new patient visits and several staff positions saved per region. No price, minimum term, volume band or billing unit is given for any of it.
The case studies report time savings without saying what the practices paid, and nothing explains how implementation and recurring charges are set for a new group, an added workflow or the embedded operations staff.
The service is built for large independent provider groups, physician owned or backed by private equity, that take referrals through several channels and run specialty scheduling. Customers span cardiology, podiatry, urology, dialysis, primary care, ENT, neurology, home health and behavioral health. ENT Partners provides a named ENT example, Upperline Health a multisite specialty group with central referral intake, and Pinnacle a fax heavy operation.
Those cases measure administrative steps, booking speed and time to enter a referral, in part of that coverage. Results for the other specialties, and for the newer voice communications product in any setting, are not reported.
Alternatives to Valerie Health
Other products that do the same kind of work as Valerie Health, and what each one does.
Medsender is a New York based company whose product reads the documents and phone calls arriving at a medical practice and files what it finds into the electronic health record.
Sold as WELL AI Inbox Admin and now as Inbox Triage within Nexus AI by WELLSTAR Technologies, a subsidiary of WELL Health Technologies (TSX: WELL), this is a fax and document intake product for Canadian clinics.
Carefeed runs admissions, workforce communication, family engagement and billing for senior living, skilled nursing and long term care organizations.
Physician-built AI that classifies, prioritizes and routes everything a practice receives from patients and referrers, and now leads with turning referrals into scheduled care.
Emitrr is a patient communication platform for independent practices that puts texting, phones, faxing, online forms, reviews and scheduling in one system, with AI agents that take calls and messages the front desk cannot.
Hello Patient (Patient Engagement Technologies, Inc.) sells AI agents, led by an assistant called Mia, that handle patient conversations by voice, text and web chat, inbound and outbound, around the clock: answering calls, booking and rescheduling, running intake, verifying insurance…
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 |
|---|---|---|---|---|
|
Contact for pricing
|
Sold through a practice demo. No billing unit, volume band or minimum term is listed. | Valerie Health acts as a business associate for covered customers. The agreement and the plan carrying it are not listed. | No implementation or onboarding charge is listed. | Vendor Published |
The demo form asks for the practice and its electronic medical record. No list prices, billing units or onboarding charges appear, and the case studies report savings without the fees behind them.