Alleviate Health

AI agents conduct the recruitment conversations that form Alleviate's main service. They personalize outreach, conduct initial study eligibility screening, and arrange visits or calls with site staff. SMS, voice prescreening with a human transfer, and voice screening followed by booking are separate workflow options. The product uses a messaging library drawn from more than 150,000 conversations and experiments with wording to improve response and conversion.

Recruiters can intervene, but their oversight surrounds an agent-led communication service. Removing the learned conversational component would remove the core automated recruitment function, leaving the site to perform the initial exchanges itself.

Recruiters can enter a participant conversation at any point. In the warm transfer workflow, the voice agent performs initial screening before passing the person to a human recruiter; other workflows book a visit or coordinator call after prescreening. The illustrative interface also shows a participant asking for more explanation and being escalated to a person. These are identifiable controls rather than a general statement that humans are involved.

Mandatory escalation criteria for distress, uncertain eligibility, or misunderstanding remain unspecified, as do confidence thresholds and the recovery process after an incorrect screening decision. Site staff retain a role without a complete set of stopping rules.

The agents use a continuously improved library of recruitment wording derived from conversation experiments. More than 150,000 conversations are associated with the messaging library, and A/B testing is intended to improve responses, callbacks, and conversion. Study setup adapts that messaging to the indication and lead source, while ongoing optimization can change the AI model and workflow.

This explains some of the operational approach, but the library is not an identified model training dataset. The language and speech model suppliers, model families, production versions, training populations, and technical approval process for updates remain unspecified.

Nylas and Twilio are named examples of connected services, and Google Workspace supports requested calendar or sign-in functions. Infrastructure providers can receive information under confidentiality agreements, while clinical trial sites receive consented information for enrollment. These disclosures identify communication and scheduling relationships, but not the language or speech model suppliers. A complete processor register, model routing details, and provider specific retention and training terms remain unspecified for participant conversations.

Lucas Research's director of patient recruitment describes faster screening and randomization after adopting the service during an influx of leads. A separate example for flu prevention recruitment among adults aged 65 or older claims a 21% improvement in voicemail callbacks. The product also claims responses within 10 seconds after a participant expresses interest.

These are concrete operational claims, but the participant counts, baseline callback rate, observation period, and evaluation design are unspecified. Customer logos establish named relationships without measuring benefit. The available results therefore do not show how performance varies by study, lead source, or participant population.

Google account information is limited to requested functions such as calendar syncing and sign-in, and is excluded from training generalized AI models. That specific restriction does not cover every participant conversation. Other service data can be used to improve functionality and AI agents, while uploaded content carries an operating and improvement license. Retention follows business purposes and legal needs without a fixed patient data schedule.

Deletion requests have a 30 day target, subject to legal obligations. Deidentification rules, broader model training limits, and technical procedures for hallucinations or AI safety incidents remain unclear. Recruiter intervention provides review during an exchange but does not settle those downstream data questions.

Alleviate carries a HIPAA compliance claim, and its May 2025 privacy policy explicitly reaches AI messaging and call agents used in trial recruitment. Covered information includes contact details, opt-in communication records, and age, location, or health conditions supplied with consent. Trial sites can receive information for legitimate enrollment purposes, while infrastructure providers operate under confidentiality agreements.

The policy separately states that SMS opt-in and consent data are not shared with third parties. This is substantive service privacy information, but explicit business associate status, a readable BAA, its applicable product scope, and corresponding subcontractor commitments remain unspecified.

Encryption, access controls, and routine audits form the stated protection for recruitment information. Account holders are also responsible for maintaining credential confidentiality and reporting unauthorized access. The privacy policy covers messaging and call agents, so these controls reach the service rather than only website browsing. An independent auditor, recognized security certification, reporting period, penetration testing scope, or retrievable assurance artifact remains unspecified.

The service connects people with clinical trial opportunities and is contractually limited to information and coordination. Its terms explicitly exclude medical advice, diagnosis, and treatment. Asking eligibility questions, scheduling a site visit, and transferring a conversation to a recruiter fit that stated recruitment role. A participant's initial prescreening result does not establish final study eligibility or replace the site's clinical assessment.

No medical device clearance is claimed for these functions. The healthcare context and collection of health information therefore do not, by themselves, turn the product's recruitment workflow into an authorized diagnostic or treatment system.

Deployment starts with study specific workflow design and messaging setup, followed by agent testing before launch. Alleviate continues to optimize messaging, the AI model, and the selected workflow until the study closes. Wording experiments target response rates, callbacks, and conversion, while recruiters retain a route to intervene in conversations. These practices provide operational testing and monitoring.

They do not establish a formal model change approval framework or comparative performance across demographic groups, indications, languages, or disability related communication needs. Release acceptance criteria, testing for missed eligible participants, and independent evaluation of model behavior remain unspecified.

Recruiters can intervene in a conversation or receive a transfer before further screening, giving the site a practical way to correct an exchange. The service terms nevertheless disclaim guarantees about the accuracy, availability, and suitability of trial information or outcomes. Liability is limited to fees paid during the preceding twelve months, and the indemnity clause runs from the user toward Alleviate.

There is no separate output accuracy warranty or contractual correction obligation for a harmed participant. A privacy deletion request addresses stored information, rather than a remedy for an incorrect recruitment or eligibility statement.

Google Workspace supports calendar syncing and secure sign-in, while the terms name Nylas and Twilio as examples of connected services. Those connections support scheduling and communication rather than demonstrating exchange of clinical records. Neither a named EHR integration nor a clinical trial management system interface is identified. The direction of patient record transfers, supported data standards, and a production clinical system connection therefore remain unspecified.

Recruiters access an online platform with accounts and connected communication services. Account holders are responsible for credentials, and the service can suspend access under its terms. Google Workspace supplies optional calendar and sign-in functions, while other infrastructure providers operate under confidentiality agreements. The hosting company, committed storage and inference regions, customer isolation design, international transfer boundaries, and private deployment options remain unspecified.

Research sites enter through a demonstration and purchase a recruitment workflow configured for their studies. Delivery includes advice on the communication approach, setup around the indication and lead sources, testing before launch, and continuing optimization through study closeout. Sites can choose SMS, voice with a staff transfer, or voice screening with booking rules, which makes the service scope understandable.

The current subscription amount, fee per call or participant, study minimum, setup charge, and limits on included support remain unspecified. The general service terms address access and liability but do not supply a price schedule for these different workflow options.

The intended users are clinical research sites recruiting and screening potential study participants. SMS is positioned for depression and neurology studies, voice transfers for situations requiring a fuller medical history, and direct voice booking for high volume studies with defined rules. A flu prevention messaging example concerns adults aged 65 or older.

Named organizations include Lucas Research, Profound Research, Alcanza, and CenExel, but those relationships do not establish equivalent performance across indications. The service is not intended for children under 18 without the parental consent conditions in its privacy terms. Comparative results by specialty, language, or participant population remain unavailable.