New · Effective 2026-09-15 · proposed at /legal/ai
1. Introduction
This statement explains how artificial intelligence is used in the WYN Agent Portal, what data those features process, what their limits are, and what protections apply. We publish it so that agencies, their clients and regulators can understand our AI systems. It supplements our Privacy Policy and Terms of Service.
2. Provider identity
Desk Plus LLC d/b/a WYN Studio, 3750 NW 87th Ave Suite 700, Doral, FL 33178. https://wynstudio.io. Contact: privacy@wynstudio.io or hello@wynstudio.io.
3. Nature and purpose of our AI systems
- Staff assistants. Halo helps agency staff plan their day and can book time blocks on their connected calendar when asked. Atlas answers questions from the agency's knowledge base and runs practice conversations. Sentinel answers staff questions about the carriers and products the agency sells and their underwriting guidelines. Argus answers questions about the agency's own records; it only reads data and proposes actions for a person to approve.
- Luna, our voice AI. Luna places outbound calls that an agency schedules, such as follow-ups on quotes and renewal reminders, and can answer inbound calls for agencies where that feature is enabled. Luna can collect information from the caller, book appointments, transfer to a person, and record a do-not-call request.
- Background assistance. Drafting emails and texts for a person to review, summarizing calls with sentiment and action items, transcribing calls and voicemail, extracting policy data from uploaded documents for a person to confirm, and scoring leads to help staff prioritize.
The purpose is to reduce repetitive work for agency staff and to respond to clients faster. No AI feature makes a decision about a person's eligibility, price or coverage.
4. Functionality and providers
- Text features use large language models from a primary provider and a fallback provider, both contractually barred from training on customer data. Sentinel runs on a conversational AI platform listed by category on our subprocessors page, under the same restriction. Agencies may instead connect their own model provider or a self-hosted model.
- Luna is a WYN Studio product built on vendor services for speech recognition, voice synthesis and language models, listed by category on our subprocessors page.
- No human voices are cloned. Luna's voice is synthetic and is not modeled on any agency staff member or client.
- Luna operates with rules for transferring to a person, ending a call, and honoring opt-outs, and is subject to per-agency controls including a kill switch, calling caps, quiet hours and do-not-call checks.
5. Human interaction disclosure
People who interact with Luna are speaking with a synthetic voice, not a person. Luna identifies itself as an AI assistant calling on behalf of the agency at the start of each call, and the website voice widget identifies itself as an AI assistant when a conversation starts. Agencies may adapt the wording but cannot remove the disclosure. A caller can ask for a person at any time. Messages drafted by AI are reviewed by a person before they are sent, except for automated notifications the agency has configured.
6. Data categories and processing
| Category | Examples | Used for |
|---|---|---|
| Audio and voice data | Call recordings, voicemail, live and stored transcripts | Running the call, summaries, records the agency keeps |
| Identifiers | Phone number, name, client record reference | Placing and linking calls and messages |
| Agency context | Policy and quote details, appointment availability, agency instructions | Making the conversation relevant |
| Documents | Declarations pages and other files an agency uploads for extraction | Producing values a person confirms |
| Usage data | Call outcomes, durations, tool actions, sentiment signals | Reporting to the agency, quality and safety |
AI providers receive only the data needed for the request and are contractually barred from using it to train models. Intermediate document-extraction results are deleted when confirmed or discarded, and in any case within 7 days. Recordings, transcripts and summaries are agency records retained under the agency's retention schedule.
7. Subprocessors and hosting
All processing occurs in the United States. Providers are listed by category at /legal/subprocessors; legal names are available on request under confidentiality.
8. Limitations and risks
- Output can be incomplete or wrong. A person must review Output before relying on it.
- Voice AI can misunderstand accents, background noise, sarcasm or unusual phrasing, and can misroute a request.
- AI features do not give legal, financial, medical or insurance advice and do not bind coverage.
- Mitigations: human review of drafts and extracted values, transfer to a person on request, per-agency controls, testing before release, and logs the agency can inspect.
9. Consent and compliance for outbound AI calls
Luna places outbound calls only for agencies that have enabled the feature and only to people the agency has authorized it to call. Before each automated call the Service checks the agency's consent records, internal do-not-call lists, the national registry data the agency maintains, the recipient's local quiet hours (8 a.m. to 9 p.m., and 8 a.m. to 8 p.m. with a limit of three attempts per day for Florida numbers), and it skips the call if any check fails. The agency remains responsible for obtaining the consent that applies to the call type, including prior express written consent for marketing calls using an artificial or prerecorded voice.
10. Rights of individuals
If you received a call or message from an agency using our Service, the agency controls your information. You may ask the agency to stop contacting you, to correct or delete your information, or to tell you what it holds. Saying that you do not want to be called during a Luna call records a do-not-call request for the agency. You may also contact privacy@wynstudio.io and we will refer your request to the agency.
11. Updates and versioning
This statement is reviewed when we add or change an AI feature and at least annually. Version 1.0, effective 2026-09-15. Changes are published at this address with a dated entry. Written in English; the English text controls.