Glossary
AI voice agent
In short
An AI voice agent is an AI system that talks with people over the phone: it listens, works out what the caller wants and replies out loud, so a call can be handled without a person on the line.
A voice agent strings three things together: speech-to-text to hear the caller, a language model to decide what to say and do, and text-to-speech to answer in a natural voice. Connected to your calendar or CRM, it can act on the conversation, not just talk.
Voice agents take incoming calls (an AI receptionist is one kind) or make outgoing ones, such as appointment reminders, confirmations and follow-ups.
The difference from older phone systems is that callers simply speak normally. There is no menu to press through and no rigid script to stay inside.
A worked example
A dental clinic gets 40 calls on a busy Monday and its front desk can take one at a time. Say a quarter of those calls are simple — opening hours, "do you take my insurance", moving an appointment. A voice agent that handles those ten calls on its own frees the front desk for the patients standing in front of it, and nobody waits on hold to ask what time the clinic closes.
When it is worth it, and when it isn't
Worth it: a large share of your calls are routine, or calls arrive when nobody can answer. Less useful: most calls are long, sensitive conversations (a complaint, a diagnosis, a negotiation) where a person is the point of the call.
Common questions
Is a voice agent the same as an AI receptionist?
An AI receptionist is one kind of voice agent: it answers incoming calls for a business. Voice agents can also make calls, such as reminders and follow-ups.
Can it make outbound calls?
Yes, but rules apply. In the US, the FCC ruled in 2024 that AI-generated voices count as "artificial" under the robocall rules, so automated outbound calls need the person's prior consent. Other countries have their own rules.
How natural does it sound?
Current voices are close to a real person; the bigger difference is speed. A good agent replies in about a second, so the call does not feel like talking to a machine.