We believe customers, regulators, and end users deserve a clear explanation of how AI is used in ReplaceReceptionist. This document is that explanation. It supplements (and does not replace) our Terms of Service and Privacy Policy.
In ReplaceReceptionist, we use AI for recognizing caller speech, understanding what the caller needs, generating spoken replies from the information you provide, and writing call summaries.
In plain English: ReplaceReceptionist is an AI phone-answering service. It answers calls to the ReplaceReceptionist number on your account (including calls you forward to it), tells every caller at the start of the call that they are speaking with an AI assistant, answers routine questions from information you provide, takes messages and booking requests (which you confirm; it never tells a caller they are booked), can transfer callers to numbers you choose while you are open, gives a scripted instruction to dial 9-1-1 when a caller describes an emergency, and sends you a summary of each call. It is not an emergency service.
Live calls run on Telnyx’s voice AI platform: Deepgram speech recognition (English and Spanish), Anthropic’s Claude Haiku model run by Telnyx (with a backup model also run by Telnyx if it is unavailable), and Telnyx text-to-speech. Call recording is off by default. We use Anthropic’s Claude models directly to read your website when you ask us to import it, for the text-only test studio, and for our backup answering engine. We do not train any model on your data or your callers’ data.
Safety rules are built into every call: the AI disclosure is spoken first and cannot be interrupted; booking requests are never described as booked; prices and estimated open times are shared only if you allow it; an emergency gets a scripted instruction to dial 9-1-1 and an instant text to you; and each transcript gets an automatic quality check that alerts us and you to a missed disclosure or a false “booked”.
Every call starts with a disclosure that the caller is speaking with an AI assistant. The example call on our website is dramatized with AI-generated voices and a fictional business.
You choose the information the assistant may use, which calls transfer to you, and whether booking requests need your confirmation (they do by default). You receive a summary of every call.
Where the Service offers an auto-send or auto-publish option, you can change that setting at any time in the dashboard. We default to human-in-the-loop wherever practical.
AI is fast and useful, but it is not perfect. Please be aware that AI Output may:
You are responsible for reviewing AI Output before relying on it, publishing it, or sending it to third parties.
AI-shopper detection works by matching what a caller says about itself (for example “I’m an automated assistant”), not by analysing voices. The test-studio mystery shop runs the same AI in a text-only simulation; no phone call is made.
Some laws require telling people when they are interacting with AI or when content is AI-generated. Here is how that works in ReplaceReceptionist today:
You are responsible for any additional disclosure required by the law that applies to you or your customers, by your industry, or by the platforms you post to.
If someone disputes a message, reply, or other output, the business that sent it can edit or remove it from its dashboard. Anyone can also e-mail [email protected] (subject line: “AI question — ReplaceReceptionist”); we will pass it to the business within one business day.
The AI processes only the data needed to perform its function. We minimize input scope wherever possible and avoid sending unnecessary personal information to the model.
We may update this Disclosure as the Service, the law, or our providers change. Material changes will be communicated by email or in-app notice at least 30 days before they take effect.
Questions: [email protected]. Legal: [email protected].
X3 E-Commerce LLC d/b/a FrontDesk Global • Michigan, United States