If you work with your hands, or you’re out with clients, the phone rings when you can’t answer it. The caller doesn’t leave a message. They ring the next business on the list.
That’s the actual cost of a missed call, and it’s why AI phone answering has moved from novelty to something worth a serious look. It’s also surrounded by more overselling than almost anything else in this field, so here’s a straight account.
What an AI receptionist actually is
A voice system that answers your phone, works out what the caller wants from normal speech, and either handles it or passes it on.
Two things separate it from what came before. Unlike a phone menu, callers just talk. There’s no pressing 1 for accounts. And unlike voicemail, it can do things while the caller is still there: check the diary, book a slot, answer a question about your hours or your rates, take proper details.
The technology got good enough recently that the pauses stopped being obvious. That single change is what moved this from irritating to usable. If you want the mechanism rather than the summary, here’s what happens between the caller speaking and the system replying.
Where it genuinely works
Answering the same questions. Are you open Saturdays. Do you service my suburb. Do you do emergency callouts. What does a standard service cost. Most businesses field the same handful endlessly, and the answers rarely change.
Booking appointments. This is where most of the value sits. Connected to your calendar, the system checks availability, books, confirms and sends a reminder. The caller never waits and you never type the booking in later.
Capturing the details that matter. A good setup asks the questions you’d ask: what’s the problem, what’s the address, how urgent is it, is this a rental or owner-occupied. You come back to something you can act on rather than a name and a number.
Out of hours. Calls at seven in the evening currently go nowhere. Answering them is often the whole business case on its own.
Overflow. When you’re already on a call, the second caller usually hangs up. Catching them costs you nothing you weren’t already losing.
Where it still falls over
Complicated or unusual requests. Anything needing judgement, negotiation or genuine problem-solving needs a person. The system should recognise this quickly and hand over rather than trying to muddle through.
Upset callers. Someone whose hot water has failed on a Sunday does not want a machine. Complaints and distress should route to a human immediately, and this is worth configuring explicitly rather than hoping.
Heavy accents and bad lines. Recognition has improved a lot, but a poor mobile connection on a windy site is still difficult. If your customers often call from noisy places, test this specifically before you commit.
Anything needing real account knowledge. Unless it’s connected to your systems, it can’t tell a caller where their order is. That connection is possible, but it’s a separate piece of work and it’s not what most off-the-shelf products give you.
The honesty question
Tell people they’re talking to an AI.
Most callers work it out anyway, and current systems are convincing enough that the discovery feels like a small deception rather than a neutral fact. The businesses doing this well simply say so at the start, and it costs them nothing. In some contexts disclosure isn’t a courtesy but a legal requirement, so it’s worth checking what applies to you.
The related temptation is giving the thing a human name and a personality. It’s a bad trade. You gain a little warmth and lose the ability to say plainly that this is an automated system that will pass you to a person when you need one.
Working out whether it’s worth it
The comparison people reach for is a receptionist’s wage, and it’s the wrong one. Most small businesses weren’t going to hire a receptionist anyway.
The right comparison is what missed calls already cost you.
Count roughly how many calls you miss in a week. Estimate what proportion would have become work. Multiply by your average job value. For a trade business with a few hundred dollars per job missing five calls a week, the annual number is usually enough to end the discussion.
Then weigh that against a monthly platform fee plus per-minute usage, which is normally modest by comparison. The maths tends to be clear-cut, which is unusual in this field and worth saying plainly.
There’s a fuller version of that calculation, including the four situations where the answer comes out as a clear no, in is an AI receptionist worth it.
What to test before you buy
The handover. Ring it yourself and give it something it can’t handle. Does it transfer cleanly, take a useful message, or loop? Looping is worse than voicemail, because the caller has already spent time getting nowhere.
Your actual questions. Not the demo script. The ten things your customers really ask, in the words they really use.
Bad conditions. From a car, from a windy site, with background noise.
The booking flow end to end. Does the appointment land in your calendar correctly, with the right detail, and does the customer get a confirmation.
What happens at 2am. Out-of-hours behaviour is often configured differently and tested by nobody.
For the capabilities to insist on rather than the tests to run, there’s a buyer’s checklist here. And if you’re wondering about the wider question behind all this, I’ve written separately on whether AI will replace receptionists.
Where the voice work sits
I built a low-latency voice prototype delivering real-time guidance, which is written up in the voice AI case study. The engineering challenge there was the same one that matters on a business phone line: responses have to arrive fast enough to feel like conversation. Half a second of delay is the difference between a system that feels natural and one people talk over.
That’s the thing to judge when you’re being sold one of these. Not the voice quality, which is uniformly good now, but the timing and the handover.
If your phone is ringing while you’re on the tools, it’s worth a conversation. The first question is simply how many calls you’re currently losing, and most people have never counted.
Frequently asked questions
What is an AI receptionist?
A voice system that answers your phone, understands what the caller wants, and either handles it or routes it. Unlike a phone menu it responds to normal speech rather than keypresses, and unlike voicemail it can book appointments, answer questions and take details while the caller is still on the line.
How much does an AI receptionist cost?
Far less than a part-time person, typically a monthly platform fee plus per-minute usage. The honest comparison isn't against a receptionist's wage though, it's against what you currently lose to missed calls. If you miss several enquiries a week, the maths is usually straightforward.
Will callers know they're talking to an AI?
Most will, and you should tell them anyway. Current voice systems sound natural but a caller paying attention can tell. Trying to disguise it damages trust when they work it out, and in some contexts disclosure is a legal requirement rather than a courtesy.
Can an AI receptionist book appointments?
Yes, and this is where most of the value sits. Connected to your calendar it can check availability, book, confirm and send a reminder, so the caller never enters a queue and you never enter the booking manually.
What happens if the AI can't handle a call?
It should hand over, and how gracefully it does that is the main thing to test before buying. Good setups transfer to a human, take a message with proper detail, or promise a callback and create the task. Poor ones loop, which is worse than voicemail because the caller has already spent time.
Wondering what this would look like in your business? A short chat is usually enough to tell.
Let’s chat