Your AI receptionist is three systems chained together, with your practice's configuration as its instructions. Knowing the pieces makes it easier to understand odd behavior and to write content the AI uses well.
The pipeline, end to end
- A caller dials your FrontDesk number and the AI answers with your greeting (unless you've set it to wait for the caller to speak first).
- Speech-to-text transcribes the caller's words as they talk.
- The language model reads the conversation plus your practice's instructions and decides what to say next, or which action to run (check availability, book, look something up, take a message, transfer).
- Text-to-speech turns the reply into audio in the voice you picked, at your speaking speed.
- The caller hears the answer, and the loop repeats for every back-and-forth (a "turn").
A reply normally starts a second or two after the caller finishes.
What's a "turn"?
A turn is one exchange: the caller says something and the AI replies.
Caller: "Hi, I'd like to book a cleaning." AI: "Sure. What day works for you?" Caller: "How about Thursday afternoon?" AI: "I have 2:30 with Dr. Park or 4:00 with Dr. Lee. Which would you prefer?"
Where your configuration plugs in
The language model knows nothing about your practice except what FrontDesk gives it on each call:
| What you configure | Where | How the AI uses it |
|---|---|---|
| Practice name | Settings → Account (organization name), or the location's Name clients see under Settings → Client Portal | The name in the greeting and the name the AI calls the practice. The location's own name is never used as the practice's name; on an account with more than one location the AI adds the office ("Bright Path Therapy, Indianapolis office") and names it when asked which office this is |
| Hours, address, holiday closures | Settings → Locations | "Are you open?" questions, directions, after-hours handling |
| Services | Settings → Services | Matching what the caller asks for to a bookable service; prices only when Quote prices is on |
| Voice, greeting, name, tone, speed | Receptionist → AI Voice → Voice | How the AI sounds and introduces itself |
| Anything else it should know | Receptionist → AI Voice → Voice | Facts added to every call |
| Knowledge base, accepted insurance | Receptionist → AI Voice → Knowledge | Looked up when a caller asks |
| What it can do, transfers, after hours, booking window | Receptionist → AI Voice → Voice | Which actions it has, and when it hands off |
| Call script | Receptionist → AI Voice → Call scripts | Extra instructions for how it handles calls |
| Clinical rules | Receptionist → AI Voice → Clinical rules | When/then rules on every call |
| Pacing and interruptions | Receptionist → AI Voice → Advanced | Turn-taking timing, silence, voicemail detection, sign-offs |
| Who answers, and when | Settings → Phone & SMS → Calls | Whether the AI, your team or a forward takes the call, day and night |
The AI is told never to make up information: if something isn't in your settings, it says so and offers to take a message. See What your AI can and can't do.
After the call
The call appears in Call Logs with a summary, transcript and recording (when recording is on), and messages the AI took appear in Tasks (and in your notifications). Each AI call is scored and added to the Review queue tab; see Review queue & clinical rules.
Why responses sometimes feel slow
- Long answers. A four-sentence reply takes longer to produce and to speak than a one-sentence one. Keep your instructions and knowledge concise.
- Actions. Checking live availability or searching the knowledge base adds a moment. Turn on Speak a filler during actions on the Advanced tab so the line isn't silent.
- Patience. A Relaxed Patience setting waits longer before replying; Snappy replies sooner.
- The phone line. A poor mobile connection on the caller's side adds delay and hurts transcription.
If most calls feel sluggish, see The AI sounds robotic or too slow.