> ## Documentation Index > Fetch the complete documentation index at: https://docs.orbit.devotel.io/llms.txt > Use this file to discover all available pages before exploring further. # Queue call-disclosure posture > Which call disclosures regulators expect by vertical, and how each one maps onto the required call disclosures you configure per queue: debt collection, recording consent, AI voice agents, and insurance # Queue call-disclosure posture A queue on Orbit can carry **Required call disclosures**: fixed statements an agent must read at the start of every call on that queue. The platform checks each call against the list live while it runs and again against the transcript after it ends. What the platform never does is choose the wording for you. This page is the companion to the operational guide: it covers the disclosures a compliance officer typically assembles per vertical, how each regime maps onto the queue phrase list, and how that list composes with recording consent, AI disclosure, and your audit evidence. If you already know your phrases and need the mechanics, go straight to [Always-on live compliance monitoring for disclosure queues](https://docs.orbit.devotel.io/guides/live-compliance-monitoring.md). ## Your phrases, your posture Orbit enforces exactly the phrases you save on each queue and ships no default disclosure rulebook. The content of each disclosure is your decision, made with your counsel: * Orbit records which phrases each queue requires and checks calls for exactly those phrases. * Orbit does not impose regulatory wording. A sample anywhere in our docs, including this page, is a starting point for counsel review, not platform-mandated text and not legal advice. * An empty list is a valid posture decision: queues without saved phrases are simply not scanned. That division is deliberate. Regulators hold you responsible for what your agents say; Orbit's job is to make your chosen wording enforceable and auditable on every call. ## Regime-to-phrase map The table below lists the regimes that most often produce a required phrase on a voice queue, what each expects your content to achieve, and sample language to take to counsel. | Regime | What regulators expect | Sample language for counsel review | Where it lives in Orbit | | - | - | - | - | | Debt collection (US) | The FDCPA "mini-Miranda" (15 U.S.C. § 1692e(11)): identify the call as an attempt to collect a debt and state that information obtained will be used for that purpose. | "This call is from a debt collector and is an attempt to collect a debt. Any information obtained will be used for that purpose." | A required phrase on every queue that handles collection calls. | | Call recording | Notice before recording begins, with consent set by the strictest participant's jurisdiction (one- or two-party). | "This call may be recorded for quality and training purposes." | A required phrase on recorded queues, paired with your recording consent mode. | | AI voice agents (EU AI Act, Art. 50) | A person must be informed they are interacting with an AI system, unless that is obvious from the circumstances. | "You are speaking with an automated voice assistant. Ask at any time and we will connect you to a person." | Your AI disclosure settings, plus a required phrase on queues where the identification must be spoken aloud. | | Insurance and annuity sales | A suitability-facing preamble: the call's purpose up front, and that the call will not verify policy or payment details. | "This call is to discuss your insurance options. We will not ask you to verify policy or payment details on this call." | A required phrase on sales queues in the vertical. | The samples are deliberately short. A phrase an agent can read verbatim in one breath is a phrase the post-call scan can find; long multi-clause recitals belong in the agent script, with the scannable core sentence configured on the queue. ## Configure the phrase list on a queue In the dashboard, open **Voice → Call Queues**, select a queue, and open its disclosure tab. Enter one disclosure per line, exactly as the agent must say it, then save. Editing is limited to owner and admin roles; everyone else sees the list read-only. * **One phrase per line.** Blank lines are dropped on parse, so formatting never becomes a phantom requirement. * **At most 25 disclosures per queue.** A counter under the editor shows the parsed count, and a 26th line blocks Save. * **At most 2,000 characters per disclosure.** An over-long line blocks Save the same way. * **Clearing the list turns monitoring off** for new calls on the queue. ### Why matching is exact Before matching, both the transcript and each phrase are normalised: Unicode composition is normalised (NFC), capitalisation is ignored, and whitespace is collapsed, so speech-to-text line breaks and double spaces do not break a match. What normalisation deliberately does not absorb is paraphrase. Matching is substring containment, not semantic similarity: an agent who rewords a configured disclosure fails the scan even when the meaning survives. Two consequences follow. * Write phrases the way agents will actually say them, and put them in the script verbatim. * When the legal text you must honour is longer than a sentence, configure the core sentence and keep the full text in the script. The scan checks the core; your QA process checks the rest. ## How enforcement composes Once a queue has a saved phrase list, monitoring needs no manual start: * **Live leg.** The moment an agent answers, the monitor listens for each required phrase and raises an alert in the agent assist panel, at most once a minute, while a phrase is still unread. The agent sees alerts while agent assist is on in the softphone. * **Post-call leg.** After the call ends and a short settling window passes, the transcript is scanned. A call where every phrase appeared is stamped with the confirmation time; a call missing a phrase pages supervisors once and fires the `call.disclosure.omitted` webhook with the call id, the missing count, and the required count. * **Escalation.** Route that webhook into your own compliance workflow: open the flagged call, listen, and decide whether it needs a callback, a file note, or a report to your compliance function. The operational details, including what agents and supervisors see, are in [Always-on live compliance monitoring for disclosure queues](https://docs.orbit.devotel.io/guides/live-compliance-monitoring.md). * **History.** Compliant stamps live on the call record, omission alerts live in the Notification Center, and the webhook feed lives in your systems. The three together are your read-back path. Queues without saved phrases, and calls that never pass through a queue, are not monitored automatically; that is a posture decision, not a gap to close. ## Pairing with recording consent and AI disclosure The phrase list covers what agents say; two neighbouring settings cover how the call itself is presented: * **Recording consent.** If any participant can be in a two-party consent jurisdiction, run recording under the announce-to-all posture described in [Call recording consent](https://docs.orbit.devotel.io/compliance/recording-consent.md) (its `announce_all` mode, seeded by `two_party_explicit`), and keep the recording notice in the queue's phrase list so the spoken announcement is scanned like any other disclosure. * **AI disclosure.** For AI voice agents, set the account-level identification behaviour in [AI disclosure settings](https://docs.orbit.devotel.io/compliance/ai-disclosure-settings.md) to meet EU AI Act Art. 50, and where the identification must be spoken on answered calls, add it as a required phrase on the queues those agents handle. The three controls are independent by design: recording consent governs the recording leg, AI disclosure governs machine identification, and the queue phrase list governs spoken regulatory statements. A queue in a regulated vertical commonly carries all three. ## Evidence for audits When a regulator or auditor asks for the trail, assemble it from the same three read-back paths: the compliance stamps on call records, the omission alert history, and the `call.disclosure.omitted` deliveries you persisted. The [evidence binder](https://docs.orbit.devotel.io/compliance/evidence-binder.md) packages Orbit-side compliance records into an export you can hand over; mirror the webhook feed from your own archive alongside it so the record stays complete for calls the platform stamped compliant. ## Limits * **No spoken revocation parsing.** The disclosure monitor checks that required phrases were said; it does not parse a caller's spoken opt-out or revocation mid-call. Spoken revocation on voice is handled by its own runbook: see [Opt-out entry points by channel](https://docs.orbit.devotel.io/compliance/opt-out-entry-points-by-channel.md). * **An aid, not a safe harbor.** A compliant stamp means your configured phrases were spoken on the call. It is evidence for your compliance program, not a legal finding and not a safe harbor; the legal weight of any wording remains with you and your counsel. ## Related * [Always-on live compliance monitoring for disclosure queues](https://docs.orbit.devotel.io/guides/live-compliance-monitoring.md) — configure, monitor, alert. * [Call recording consent](https://docs.orbit.devotel.io/compliance/recording-consent.md) — one- and two-party modes for the recording leg. * [AI disclosure settings](https://docs.orbit.devotel.io/compliance/ai-disclosure-settings.md) — machine identification under EU AI Act Art. 50. * [Restricted & Prohibited Industries](https://docs.orbit.devotel.io/compliance/restricted-industries.md) — what you may send and say per vertical. * [Evidence binder](https://docs.orbit.devotel.io/compliance/evidence-binder.md) — auditor-ready exports. * [Webhook events](https://docs.orbit.devotel.io/webhooks/events.md) — the `call.disclosure.omitted` payload. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.