How the AI receptionist works
ARIA answers the phone, works from a knowledge base you write, collects the details you specify from each caller, and transfers to the right person when a human is needed. Every call is transcribed, so nothing depends on remembering what was said.
Setting one up
An assistant is configured, not trained. Everything it knows is something you entered.
-
Give it an identity
A name and an opening line. Separate open and closed messages mean a caller after hours hears something different from a caller at eleven in the morning, without a second assistant.
-
Fill the knowledge base
Hours, services, directions, policies, anything a caller asks. This is the assistant's only source of truth, which is deliberate: when an answer is wrong, you fix the entry rather than arguing with a model.
-
Say what to collect
Define the fields the assistant should capture from each caller, such as name, date of birth, reason for calling and a callback number. It works through them in conversation.
-
Set the transfers
Name the numbers for the front desk, billing or on-call. When a caller needs a person, they reach the right one instead of a queue, and the transfer is flagged in the transcript.
More than one assistant, doing different jobs
A practice is not one phone tree. Each assistant has its own number, greeting, knowledge base and transfer targets.
- Separate assistants for the main line, an after-hours line, or a specific service
- Each has its own phone number and its own configuration
- Voice and language are chosen per assistant
- The same configuration drives an optional chat widget on your website, so a visitor asks the same questions and gets the same answers
- Reports hold the call analytics and the transcripts
Questions we get about the AI receptionist
What happens when ARIA does not know something?
It answers from the knowledge base you wrote, and hands off to a person when a caller needs one. If an answer is wrong, the fix is to correct the knowledge base entry, because that is the only place it draws from.
Can callers still reach a human?
Yes, and that is the point of transfer targets. You name the numbers for the front desk, billing or on-call, and transfers are flagged in the call transcript so you can see how often it happens.
Can it say something different after hours?
Yes. Open and closed messages are configured separately and swap automatically, so an after-hours caller is not told to hold for a desk that is empty.
What details does it collect from a caller?
Whatever you define. Name, date of birth, reason for calling and a callback number are typical, and the assistant works through the fields you set in conversation.
Where do we read what callers actually said?
Every conversation is transcribed and available in the reports, and recent ones surface on the main dashboard. Transfers are marked in the transcript.
Can we run more than one assistant?
Yes. Each has its own phone number, greeting, knowledge base, voice and transfer targets, so a main line and an after-hours line can behave completely differently.
Does this work on our website too?
Yes. The same assistant can be embedded as a chat widget, so a site visitor asks the same questions and gets answers from the same knowledge base.
How do we try it before committing?
Assistants can be provisioned as a pilot with its own link, then migrated to a full account if it works out, rather than switching the main line over on day one.
Hear it answer your own phone questions
We will load a few of your real practice details into a knowledge base and let you call it.
Talk to our team