What does an AI receptionist for trades actually do?
An AI receptionist for trades answers the phone, web chat and social messages you cannot get to, works out what the job is and whether it is worth your time, quotes from your real price list where the job is standard, and books it straight into your calendar. Everything it says is logged and searchable, so you can read every conversation it had while you were under a house.
It is not a menu tree and it is not a chatbot on a website. It talks like a person, it holds a conversation across fourteen or fifteen turns, and it knows your services because it was built from them.
Why does a missed call cost so much in a trade business?
Because the person calling you is not waiting. They have a burst pipe, a dead hot water system or a fence blown flat, and the third name in the search results answers on the second ring. In most trades, an enquiry you do not answer inside a few minutes is an enquiry that has already become someone else's job.
The maths is unkind and simple. Take the number of calls you miss in a normal week, take the proportion that would have become work, and multiply by what an average job is worth to you. Most owners we sit with have never done that sum, and most of them go quiet when they do.
That number is also the only sensible way to decide whether any of this is worth buying. If a machine cannot beat it, no price is cheap.
When should it answer, and when should you?
It should answer when you cannot, which for most trades means evenings, weekends, and any hour you have both hands occupied. You should answer everything else.
We do not think an AI should replace the call you would have happily taken. Your voice on a first call is worth more than any machine, and a customer who wanted you and got a robot remembers it. The value is entirely in the calls that currently ring out.
Start it after hours only. Watch what it catches for a month. Widen it if, and only if, the log says it deserves more.
What should an AI receptionist never be allowed to do?
It should never invent a price, never promise a time you cannot make, and never handle an emergency without a clear escalation path to a human.
Those three rules cover most of the horror stories. A machine that quotes outside your real price list will eventually quote something that loses you money on a job you are contractually stuck with. A machine that books a slot your van cannot reach turns an enquiry into a complaint. And a machine that keeps chatting pleasantly while someone describes a gas smell is a genuine hazard.
The correct build has your prices as hard boundaries, your real availability as the calendar it books into, and a defined set of words and situations that trigger an immediate handover. Ours also asks before it acts for as long as you want it to, because approval gates by default is a standing commitment here rather than a feature we sell back to you.
Will customers know they are talking to a machine?
They will, and they should be told outright in the first few seconds of the call rather than left to work it out.
There is a temptation to build something that passes for human and never admits it. Resist it. Australian customers are not offended by an assistant that says it is an assistant, takes their details properly and gets them a call back at eight tomorrow morning. They are extremely offended by finding out later that they were fooled, and that feeling attaches to your brand, not to the software.
The honest version also performs better in practice. People give clearer information when they know they are talking to a system that is writing it down.
What does it cost, and how long does it take?
Our version of this is $1,950 to set up and $295 a month to run, and it is typically live in two weeks. That is the whole published figure, setup and running, and it is on the page rather than behind an email form.
The monthly fee is the part worth understanding. Your prices change, your services change, and somebody eventually adds a new offering that nobody tells the machine about. The run fee covers monitoring and tuning so that a build still behaves in eighteen months. A machine left unattended is not an asset.
One thing to insist on with any supplier: a clear fair-use cap and a published overage rate, or the machine running on your own provider account so you see the usage bill directly. Call volume in a storm week is not the same as call volume in a quiet fortnight, and you should know the ceiling before you sign.
How do you know it is working?
You agree the numbers before it is built, then you read them thirty days after go-live.
For a lead machine the numbers are usually plain: enquiries answered outside business hours, jobs booked without you touching the phone, and the response time on a first contact. Thirty days after go-live you should get one page showing those numbers against target, in English. If it did not earn its keep, that should be a conversation about stopping, not a conversation about upgrading.
You can also just read the transcripts. Every conversation is logged, which means the honest evidence is sitting there whether it flatters the build or not.
Where to start
If you want to see one working before you talk to anybody, Keeper is playable on a staged plumbing business: text it like a customer and watch it qualify and book. The full picture of what it does in a real business is on the never miss a lead page, and Keeper's own page walks through what you provide, what it does and what you get back.
Worth reading alongside this: the Australian Government's AI Ethics Principles on transparency, and the Australian Privacy Principles, since you are about to start recording customer conversations at scale.
When you are ready for a number, tell us what the missed calls are costing you and we will tell you what it costs to stop.