Hello Patient

Conversational agents are the product. Hello Patient's agents answer calls, texts and web chat, book appointments, run intake, answer insurance questions, take refill requests, run recall and billing outreach, and, since its September 2026 acquisition of Converse Health, read incoming faxes, file documents into the record, find overdue follow-ups in the chart and draft authorization paperwork.

The company says it builds on existing foundation and voice models, fine-tuned for healthcare front and back office work, with a multi-agent design for complex workflows. Take the agents away and what remains is integrations and a phone line, not a product a practice would buy.

The rules that send a patient to a person are published by category, and the agent's own share of clinical routing is left to each customer. Hello Patient's FAQ says the agent follows the rules the practice sets, answers approved questions, and escalates dosing changes, side-effect concerns, prescription issues and anything needing clinical judgment to a person on the team; on urgent care lines, a question about symptom severity or whether to come in or go to the emergency department is escalated to a person 'or directs the patient per the protocols you set'.

For authorization paperwork the agent composes the request, routes it to staff for review and does not submit anything to a payer. Every interaction is logged for review, and the agents cover every call, text and chat around the clock. Nothing says which protocols let the agent direct a patient itself rather than transfer, or what the patient hears while waiting for a person.

The build approach is stated, and the models are not named. Hello Patient says it does not build its own foundation or voice models; it fine-tunes existing ones for healthcare front office work, has been fully generative and conversational from the start rather than layering AI onto scripted flows, and uses a multi-agent architecture for complex workflows. It names its voice infrastructure partner. Which foundation and voice models sit underneath, how the fine-tuning corpus was built, and how accuracy is evaluated are not published; containment is the only performance figure given.

One party in the chain is named, and the model layer is not. Hello Patient names its voice infrastructure partner and says it builds on existing foundation and voice models rather than its own, so third-party model providers sit in the call path without being identified.

Its privacy policy says it holds business associate agreements with its third-party service providers and that its servers are in the United States, and its FAQ says protected health information stays inside its environment and is not used to train AI models. No subprocessor list is published, and nothing names the telephony, speech recognition, speech synthesis or foundation model providers, or the retention terms they operate under.

Named deployments with periods and denominators, reported by the vendor. Hello Patient's Pulse-MD case reports the missed-call rate across four urgent care locations falling from 26 percent to zero in the first 30 days, with 30 percent containment (calls finished without a person) and 208 staff hours returned.

For Found Health, a consumer health brand, the vendor reports a 140 percent lift in conversions over its previous one way text reengagement, recovering 4,739 stalled signups across a 92,290 lead program over 16 months. The home page adds unnamed averages: 100 percent answer rates, 18 percent more appointments booked and a 95 percent reduction in wait times, with 1.3 million conversations handled.

Across consumer health programs agents are said to handle 70 to 80 percent of routine inbound questions. None is independently measured, and nothing reports the accuracy of what the agent tells patients.

A categorical training commitment is published, without a retention schedule. Hello Patient's FAQ says it does not use protected health information to train AI models, that protected health information stays inside Hello Patient's environment and is used only to serve the practice that owns it, and that every conversation is logged and encrypted. Its privacy policy keeps personal information as long as necessary for its purposes or as the law requires.

No retention period is published for call recordings, transcripts, faxes or chart extracts, and nothing says whether de-identified conversations are used to tune models. The data now includes referral faxes and chart notes as well as calls. Safety engineering, meaning guardrails, hallucination mitigation and how a safety event is handled, is described only through approved-content rules and escalation.

Business associate status is stated in a privacy document that reaches the service, and the agreement is not published. Hello Patient's FAQ and its September 2026 acquisition announcement say it is HIPAA compliant and signs a business associate agreement with every client, for back-office document work as well as conversations.

Its privacy policy, issued by Patient Engagement Technologies, Inc., has a HIPAA section saying it holds business associate agreements with its healthcare provider partners and its third-party service providers and uses protected health information only as permitted. The agreement text, breach notice period and the list of service providers it binds are not published. Veterinary practices use the same platform outside HIPAA.

A SOC 2 Type 2 attestation is named as the company's own, with nothing retrievable behind it. Hello Patient states SOC 2 Type 2 certification on its site and in its FAQ, alongside HIPAA compliance, and its privacy policy describes regular security assessments. No report period, auditor, trust center or request route is published, and no other certification is named.

Hello Patient says its agents never give medical advice, and its own FAQ describes them answering and routing clinical questions for patients. On consumer health programs the FAQ says the agent improves conversion by answering the questions automation cannot, including cost, eligibility and dosing, and answers approved questions about eligibility and common member concerns.

On urgent care lines, when a patient asks about symptom severity or whether to come in or go to the emergency department, the agent escalates to a person 'or directs the patient per the protocols you set'. An answer about dosing, or a direction to the emergency department or to stay home, reaches the patient with no one on the line, and the customer's protocols are neither named nor published. Nothing shows the agent reading the practice's approved text word for word with everything else transferred, and no regulatory position is stated.

Deployment testing is described, and no governance framework or results are. Hello Patient's FAQ says each go-live runs through protocol mapping and content tuning with the customer's staff, integration and conversation testing, and a pilot with real patients before launch, and that every interaction is logged for review. No governance framework, model monitoring program, error analysis or bias evaluation is published.

The agents support English and Spanish, with other languages on request, and no performance is reported by language or accent, which matters for a system whose typical failure is mishearing. Containment figures are now published for some deployments, but nothing reports how often an answer was wrong.

Mechanisms let a practice review the agent's work, and nothing stands behind a wrong outcome. Every conversation is logged for review, clinical questions outside the practice's rules go to a person, and authorization requests are routed to staff for review rather than submitted by the agent. No warranty, remedy, service credit tied to outcomes or error rate is published; the containment figures measure how much the agent finishes, not how often it is right. The costly errors are a patient told to stay home who needed care, a wrong dosing or eligibility answer on a consumer health line, and a referral fax filed to the wrong patient's chart.

Named record systems with read and write described for one of them. Hello Patient's FAQ lists record, practice management and CRM systems it has worked with, including AdvancedMD, athenahealth, eClinicalWorks, ModMed, NextGen, Provet Cloud, Salesforce and Zenoti, and supports FHIR and HL7, with CSV or custom work for systems without API access.

For athenaOne it describes a native integration that reads patient records and calendars and writes new appointments, reschedules, cancellations, new patient accounts and appointment notes, used by Privia Health practices. The Converse Health agents also file faxed documents into the record. No marketplace listing or published integration documentation is cited.

Residency is stated, and the call path is left open. Hello Patient's privacy policy says its servers are located in the United States and that data from outside the country is transferred to, stored and processed there, and its FAQ says protected health information stays inside Hello Patient's environment. It is a single vendor-hosted service reached through integrations.

The company names no hosting provider or tenancy model and does not say where telephony, speech recognition, speech synthesis and model processing happen, although each of those steps handles call audio or transcripts. Veterinary and human health customers share the platform.

No price is published, and the pricing basis is. Hello Patient's FAQ says it prices on a usage-based model that scales with conversation volume across text, voice and web chat, with the cost depending on volume, the mix of inbound and outreach, and how much the agent handles versus existing workflows; buyers bring their numbers to a call for a model. Privia Health practices get a network discount. No rate, minimum commitment or implementation fee is published, and go-live typically takes about six weeks.

Broad across outpatient settings and consumer health, with named validation in urgent care and a direct-to-consumer brand. Hello Patient sells to multi-location medical groups in urgent care, primary care, dermatology, orthopedics, ENT, dentistry, optometry and mental health, to Privia Health practices as an approved vendor, to consumer and telehealth brands, and to veterinary practices on the same platform.

Pulse-MD and Found Health are named with results; inpatient and acute care are out of scope. Language support is English and Spanish. The work itself is patient access and document intake: the agents answer and route calls, book and reschedule, run intake and verify insurance, and since September 2026 read referral faxes, match or create the patient record and file documents into the chart.