Basata
Basata builds AI agents for the administrative work of specialty practices: referrals, faxes, intake, records requests, refills, scheduling and phone calls. Document agents read each incoming fax, sort it as a new referral, a result for an existing patient or a records request, pull out the referral details and create the patient's chart in the EHR. When a referral is missing information, the software faxes the referring office for it. For a records request it assembles the documents, and staff approve the release with one click. Voice agents answer calls, take refill requests, confirm insurance and demographics and send a structured message to the right nurse or provider, and an interaction outside an agent's role goes to a person.
Basata raised a $21 million Series A led by Basis Set Ventures in May 2026 and reports more than 500,000 patients served across cardiology, urology, gastroenterology and ophthalmology groups. Southwest Cardiovascular Associates cleared a backlog of more than 500 referrals, moved from four to eight weeks of delay to same day processing and booked 18% more new patients. Midwest Cardiovascular Institute runs a voice agent called Ava at 1% call abandonment. Deployments are scoped and quoted after a 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
Every product here is an agent: a document model that classifies faxes, matches them to a patient and extracts referral fields, and voice agents that hold the call, confirm identity and insurance and write the message that reaches a nurse. The surrounding pieces, an analytics dashboard, fax and record connections and the review queues, exist to carry agent output into the practice. Remove the models and nothing is left to sell, since the referral backlog, the call queue and chart creation all return to staff.
Agents work inside assigned roles with published limits: voice agents may not give medical advice or clinical direction, must confirm identity when required and pass any interaction outside their role to a person, and a patient can ask for a human at any time. Document output is staged in review queues, refill requests go to a provider for sign off, and staff approve each records release. Missing are the confidence level at which an extraction is held back for review and what happens once a staged chart or routed message proves wrong.
The stack is described by function rather than by name. A document layer reads unstructured faxes, lab orders, pharmacy documents and referrals, extracts fields and applies the practice's routing rules, while conversational voice and chat agents run on the practice's own scheduling logic. Outside AI providers process documents for individual tasks. Those models, and any version or update practice, go unidentified, so a buyer cannot tell when the system behind a voice agent or an extraction changed.
Documents can go to outside AI providers for analysis. Those providers must meet Basata's privacy and security requirements, sign a business associate agreement where they may touch protected information and use the data once for the task without keeping it, and partners are reviewed on a recurring basis. Model, speech, fax and hosting suppliers go unnamed and there is no subprocessor list, so a buyer cannot map which outside services receive a faxed referral on its way to the chart.
Southwest Cardiovascular Associates, in a case study dated May 2026, went from a backlog of more than 500 referrals and four to eight weeks of delay to same day processing, booked 18% more new patients and absorbed 50% provider growth without adding fax staff, against at least 20 minutes a fax by hand. N.E.O.
Urology Associates reports 3,000 calls a month answered on the first ring and four times the staff capacity, Phoenix Heart six times faster fax handling, and Midwest Cardiovascular Institute 1% abandonment with holds under 30 seconds. The figures come from the practices and Basata, with no independent measurement.
Customer data is never used to train models, and an outside AI provider uses a submitted document once for the task without storing it afterward. Agents receive only the patient information, systems and tools their task requires, and persistent information is written to the provider's own systems under the organization's retention rules. How long Basata keeps call recordings, transcripts and staged work in its own platform carries no stated window, and guardrails against a voice agent inventing an answer are not described.
Basata calls itself HIPAA compliant with administrative, technical and physical safeguards built in, and every third party that may process protected health information signs a business associate agreement and is reviewed on a schedule. Its own business associate role toward a practice goes unstated, and the Privacy Statement and Terms links in the site footer lead nowhere, so no document reaching the service supports the compliance claim. A practice settles its agreement, and what that agreement covers, in contracting.
Basata states it is SOC 2 Type II attested and shows the AICPA mark across the site, alongside ongoing monitoring, testing and alerting, agent access limited by role and timestamped histories of every agent action. The report, its audit period and whether its scope reaches the voice and document agents and the outside AI providers have to be requested, since there is no trust portal and nothing to download.
The agents are built to support healthcare administration rather than practice medicine: referral intake, chart indexing, records release, refill requests, scheduling and call routing in specialty practices. Voice agents are restricted from giving medical advice or clinical direction, and refill requests pass to a provider for sign off. What is sold matches that scope and sits outside device regulation, though no specific exclusion is cited.
Each agent goes through a test suite before deployment, adversarial tests for unexpected behavior included, with checks that it follows instructions, protects patient information, confirms identity and escalates, and live agents are monitored continuously. These are commitments without results: no evaluation outcome, error rate or named framework accompanies them, and the security principles article describes good practice in general terms rather than how these agents performed. Nothing shows how the voice agents fare with callers of different accents or languages.
Staff can check agent work before it takes effect: staged charts sit in review queues, records releases need a click of approval, and document and call histories show what an agent received, extracted, routed and changed. Nothing stands behind the output itself. The footer's terms link is a placeholder, so before a contract there is no warranty, remedy or correction route for a misrouted referral or a wrong chart entry.
Agents write into the record: they create new patient charts from referrals, index results and notes into existing charts and assemble records for release. The NextGen integration stages documents in dedicated review queues for staff to verify before scheduling, and the company says it connects to all EHR and fax systems. NextGen is the only record system named, and no marketplace listing or interface documentation shows what is read and written in other systems.
A hosted service reaches the practice through defined connections to its record, fax and phone systems, and persistent information is written back to the provider's own systems under its retention rules. Call audio, documents and extractions pass through Basata and its AI providers on the way, with no hosting provider, country or region committed for them and no self hosted option.
A demo request is the only route in. Deployments are scoped workflow by workflow, often starting with voice or referral intake and expanding from there, with Basata staff spending days on site to map how the practice works. No price, billing unit or implementation charge appears anywhere, and nothing indicates whether cost follows documents, calls, providers or sites.
Independent specialty groups and their management organizations are the market, in cardiology, urology, gastroenterology and ophthalmology, across front office intake, referral processing and call centers. Case studies and testimonials back cardiology, at Southwest Cardiovascular Associates, Midwest Cardiovascular Institute, Phoenix Heart and Tri City Cardiology, and urology at N.E.O. Urology Associates. Gastroenterology and ophthalmology are named without deployment results so far, and the agents stay out of clinical decisions.
Alternatives to Basata
Other products that do the same kind of work as Basata, and what each one does.
Valerie Health runs an AI front office for independent physician groups, combining software agents with its own operations staff.
Aidin runs the discharge referral, from the point a case manager decides a patient needs post acute care to the point a facility accepts them.
ReferralMD runs referral management for practices, health systems and provider networks, bringing faxed referrals, online submissions and record feeds into one work queue.
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…
Parakeet Health gives provider groups AI agents for scheduling and routine patient communication. A practice can connect its phone lines and other channels and configure how an agent handles appointment types, locations, eligibility questions and individual provider preferences.
ArcHouse, made by HospiceMatch, Inc., is software for the work between a referral and an admission at hospice, palliative care, home health and house call providers.
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 workflow quote
|
Provider organizations purchase through a sales engagement scoped to the records, fax, call or scheduling workflows selected. The charges and billing units are agreed for the deployment. | Subprocessors handling protected health information must sign business associate agreements. The customer's agreement is handled in contracting. | Configuration and integration depend on the workflow and the systems used by the organization. | Vendor Published |
The price depends on the configured tasks and systems. The rate, billing unit and scope of engineering services are agreed with the provider.