An AI call answerer is a phone system that talks. It picks up when you can't, asks the customer questions, and passes what it learns to you. Underneath, it's an old idea with a newer voice.
The old idea underneath
Phone systems that answer callers aren't new. Interactive voice response, or IVR, describes automated telephony systems that interact with callers, gather information, and route calls to the appropriate recipient1. The classic version says "press one for sales."
An AI answerer does the same job but lets the caller just talk. Three parts are doing the work:
- Speech recognition turns the caller's voice into text.
- A language model reads that text, decides what to ask or say next, and keeps the thread of the conversation.
- Text-to-speech turns its reply back into a spoken voice. IVR systems have long used text-to-speech to speak complex and dynamic information1.
Where it goes wrong
The first part is the shakiest. Speech varies in accent, pronunciation, pitch, volume and speed, and background noise, echoes and recording quality distort it2. A plumbing call is often made from a bathroom floor or a car, with a tap running. Street names and surnames are the likeliest to be misheard.
A good setup repeats the key details back to the caller. "That's Wattle Street, spelled W-A-T-T-L-E?" costs five seconds and saves a wasted trip.
What it needs from you
The answerer knows nothing about your business until you tell it. Expect to supply:
- Your services and areas. What you do, what you don't, and which suburbs or postcodes you cover.
- Your hours. When a person is available and when the answerer is on its own.
- The questions to ask. The list of job details worth collecting is a starting point.
- A way to reach you. Calls that need a person have to go somewhere. Phone services can connect the current caller to another party, but if the line is busy, there is no answer, or the called party hangs up, the call ends there unless you've planned a fallback3.
- Where bookings go. A text, an email, or your job software. See connecting AI to job and calendar software.
Tell callers what they're talking to
In Australia, the privacy regulator says public-facing AI tools such as chatbots should be clearly identified as such to users4. A voice that sounds human but isn't should say so near the start. Check the rules where you operate, and read the regulator's page for your country rather than relying on a summary here.
The same guidance talks about building human oversight into processes4. For a small plumbing business, that can be as simple as reading each day's call summaries and listening to a few recordings, where recordings are allowed.
What it won't do
It won't know your van is stuck in traffic. It won't decide whether a job is worth taking. It won't fix a diary that's already overbooked. It takes calls, collects information and passes it on. Everything after that still needs a person who makes decisions.
Where the facts come from
- Interactive voice response, Wikipedia
- Speech recognition, Wikipedia
- TwiML Voice: Dial, Twilio
- Guidance on privacy and the use of commercially available AI products, Office of the Australian Information Commissioner (OAIC)