Start with the maths that sells this use case by itself. Service businesses miss a startling share of their inbound calls, on the tools, after hours, already on the phone, and a caller who rings out mostly does the obvious thing: rings your competitor. If your average job is worth even a few hundred dollars, a handful of rescued calls a week is a five figure annual difference, before you count the after hours enquiries that currently go straight to voicemail purgatory.
This is why voice agents are usually the first thing I look at for any business that lives on its phone.
What a modern voice agent actually does
Forget the phone robots you have suffered through, "press one for". A modern voice agent is an LLM wearing a headset: it answers instantly, in a natural voice, and holds an actual conversation.
- Answers every call, first ring, around the clock, including the 7pm enquiry and the Saturday emergency
- Handles the routine questions from your FAQ: do you service my suburb, what do you charge for X, are you taking new patients
- Books appointments straight into your real calendar, within rules you set
- Takes a proper structured message when it cannot help: name, number, address, the issue in the caller's own words
- Recognises the hot lead or the emergency and escalates immediately, transferring to a human or triggering an instant text to you
- Logs every call as a transcript and summary, so nothing lives only in someone's memory of a phone call
What it should not do
The boundary design matters more than the voice quality. My rules: it discloses that it is an AI assistant when asked, and I favour being upfront regardless, because trust is worth more than a seamless illusion. It does not quote complex jobs, price on the fly, or give advice in regulated domains, it captures and escalates. And an angry or distressed caller goes to a human fast, per the trust moments principle from the previous article. A voice agent that knows its limits builds your reputation. One that bluffs past them spends it.
What to prepare before you deploy one
The technology is the easy half. The preparation is a one afternoon job and it determines everything:
- Your twenty most common questions, with the answers you actually want given
- Your booking rules: durations, buffers, areas served, what never gets booked without a human look
- Your escalation triggers: the words, situations, and customer names that mean get a person now
- Where the outputs land, calendar, job system, inbox, so calls become workflow instead of sticky notes
If you cannot produce that list, that is the finding: the phones are currently running on tribal knowledge, and writing it down pays even before any AI touches it.
Where to start
Do not begin by replacing your main line reception. Begin where the calls are currently dying: after hours, overflow when the line is busy, or a second number on your Google listing. Zero risk to the experience you already deliver, pure rescue of the calls you were losing, and a fortnight of transcripts will tell you exactly how far to extend it.
Video: a live call with an AI voice agent, and the booking it created
Phones sorted, the next fire hose is the inbox. That is the next article.