Water does not wait for business hours. That single fact is why plumbing has a harder phone problem than almost any other trade, and why voicemail is more expensive here than it is anywhere else.
Someone standing in an inch of water at 11pm is not leaving a message. They are working down the search results until a human voice answers. Whoever picks up gets a job at emergency rates from a customer who is not price-shopping.
Reign AI builds AI receptionists for trades — human-grade voice AI that answers at 11pm on a Sunday — so we have a stake in what follows. The numbers below are still yours to run first, and for plenty of plumbing businesses they point at a better voicemail greeting instead.
The shape of a plumbing phone book
A real share of your calls are genuine emergencies. Burst pipes, backed-up sewage, no heating in February. These callers convert at rates the rest of your marketing cannot touch, and they convert to whoever answers first.
The rest are highly repetitive. Blocked drain, dripping tap, running toilet, low pressure, boiler quote. The same twenty questions, asked over and over, none of which need you personally.
You are physically unable to answer. Under a sink, arms wet, in a crawlspace. There is no version of the job where the phone is convenient.
Some of them are not your job at all. A property on a septic system with sewage surfacing outside needs a different trade, and the caller cannot tell the difference — the questions that sort it are set out on the septic and excavation page. Asking them costs one minute; not asking them costs a call-out.
Callers self-triage badly. “It’s leaking a bit” is sometimes a dripping washer and sometimes a failing joint about to let go. A good agent asks the follow-up questions that separate those; voicemail asks nothing at all. A human answering service will take the message but generally will not triage it either — that trade-off is set out in AI receptionist vs answering service.
What it handles well
The 2am emergency
This is the highest-value case, so it is worth spelling out the actual sequence — the general mechanics, for any trade, are in how AI receptionists work. The call is answered on the first ring. The agent establishes what is happening, and — if you configure it to — gives the one piece of immediate advice that saves the customer thousands: find the stop tap and shut the water off. It captures the address, gets a callback number, establishes whether the water is contained, and escalates to you within seconds.
That advice matters. The difference between a customer who shut the water off in the first minute and one who waited forty minutes for a callback is often the difference between a repair and an insurance claim. It costs nothing to configure and it is the sort of thing customers remember.
Routine job intake, done consistently
For everything non-urgent, the agent works through the questions you actually need: what the problem is, how long it has been happening, property type, whether they own or rent, whether the water is currently on, and whether they have tried anything already. Then it books the visit or captures the details.
The value is not that it asks better questions than you — it is that it asks the same questions every time. Nobody’s memory does that on call number nine of the day.
Screening the work you do not want
If you do not do commercial, do not touch gas, or will not travel past a certain point, the agent declines politely and stops. Every one of those calls avoided is time you did not spend on a callback to say no.
After-hours and overflow
If you already have someone on the phones, the highest-value configuration is often not replacing them. Point the overflow at it — when the line is busy, and everything outside business hours. Your daytime calls carry on exactly as they do now, and the calls that used to hit voicemail turn into booked jobs and flagged emergencies.
A sample call
An illustrative example, not a recording of a real call. It assumes a business in Ontario that has asked for shut-off advice to be added to its receptionist’s instructions, and has not switched on live transfer.
Receptionist: Thanks for calling [Company name]. This call may be recorded for quality and training purposes. How can I help you?
Caller: A pipe froze in the basement last night and now it’s burst. There’s water spraying everywhere.
Receptionist: Okay. First, do you know where your main water shut-off valve is? It’s usually where the water line comes into the house, near the meter.
Caller: Hang on… got it. It’s off. The spraying’s stopped.
Receptionist: Good — that’s the most important thing. Is any of the water near electrical outlets or the panel?
Caller: No, it’s on the floor by the laundry sink. Am I talking to a real person?
Receptionist: No, I’m an AI receptionist for [Company name]. I’ll make sure the team has everything they need. What’s the address?
Caller: 88 Elm Street, Mississauga. It’s Priya. (Gives a callback number.)
Receptionist: Thank you. I’ve noted a burst pipe in the basement with the main shut off and no water near anything electrical.
Caller: How much is this going to cost me?
Receptionist: I can’t price a repair without someone seeing it — the plumber will go through that with you. Someone will call you back as soon as possible.
The owner gets a summary by email, or by Telegram if it has been connected: the name, address and callback number, a burst pipe with the main now off, water on the floor away from the electrics, and the price question. The receptionist did not promise an arrival time, quote a figure, or read a number back off caller ID — it asked. A business that has switched on emergency transfer would hear one more question first — is this an emergency right now, or can it wait until the next business day? — and on a clear yes the receptionist rings the on-call plumber before taking any details, coming back to take the message if nobody answers.
Two things in that call reflect a Canadian business. Every caller hears the recording notice in the greeting, and in Canada the receptionist says it is an AI whenever a caller asks rather than announcing it up front — a US business’s greeting says so at the start. And the scenario itself follows the season: a pipe that freezes overnight often gives way as it thaws, so burst-pipe calls bunch up in the days after a cold snap rather than arriving evenly through the winter. The same setup serves a business in Canada or the US, and the number on the van keeps working — calls are forwarded to the receptionist rather than moved to a new number. The way to judge any option is a live test call — the checklist is at the end of this page, and the mechanics underneath are in how AI receptionists work.
A worked example — the after-hours arithmetic
The method: count the after-hours calls, decide how many are genuine emergencies, compare what you win from them now against what you would win answering live, and price the difference. That comparison is the transferable part. The figures below are assumed — a worked example, not data about your business or a research finding — so substitute your own.
Assume a small plumbing outfit:
- 6 after-hours calls a week currently going to voicemail
- 2 of those are genuine emergencies
- You currently win 1 of the 6, from a voicemail returned the next morning
- With every call answered live, say you win 3
That is 2 additional jobs a week. At an after-hours average of $600, that is $1,200 a week — around $62,000 a year — from calls that were already dialling your number.
Even if you halve every assumption, the answer does not change much. The reason to actually run the numbers rather than accept the pitch is the opposite case: if you get one after-hours call a month, none of this is worth doing, and you should know that before you spend anything.
Your call log has the real figure. It is usually worse than the guess — the cost of missed calls in plumbing walks through where to find each input and what the total actually means.
If the out-of-hours window is the whole of your problem rather than daytime overflow, the options are compared side by side in after-hours answering for contractors.
Configuration: the part that decides whether this works
Your emergency criteria. Be specific. A workable starting set: uncontrolled water, no water at all, sewage backing up, no heat in winter, a leak near electrics, gas smell (which should route to the gas emergency line, not to you). Then decide what happens for each — text, ring, transfer live.
Your after-hours promise, honestly. If your real answer is “we’ll be there within two hours” then say that. If it is “we’ll ring you back first thing”, say that instead. The most common way these deployments go wrong is an agent promising a response time nobody keeps. That converts a win into a complaint, and the customer blames you, not the software.
Your pricing policy. A fixed callout fee is a real number and the agent can state it. A boiler replacement is not, and the agent should say pricing follows an assessment. Decide which of your work has a phone-quotable number and which does not, and make the agent firm about the boundary.
Gas work, explicitly. If you are not gas-qualified, the agent must be unambiguous about it. And a caller reporting a gas smell should always be directed to the emergency gas line first — that is a safety rule, not a scheduling preference, and it should be configured before anything else.
Immediate advice. The stop tap. Whether to turn off the boiler. Anything else you would tell a panicking customer in the first thirty seconds. This is free goodwill and most setups never bother.
Service area and property types. Precise enough to turn down a caller outside it cleanly.
What it will not do
Diagnose. “Why is my pressure low” has half a dozen causes and needs someone looking at the system. The agent’s job is to capture the symptom accurately, not to explain it.
Price complex work. Boiler swaps, bathroom refits, re-pipes. Any agent confidently pricing these over the phone is creating a dispute for later.
Replace judgement on a genuinely ambiguous emergency. It applies your rules. When the description does not clearly fit them, the right behaviour is to escalate to you and let a person decide — and you should verify it does that before you rely on it.
Handle an angry customer about existing work. A complaint about a job you already did needs you, and the agent should route it to you rather than attempting to smooth it over.
Test it before you trust it
Ring your own agent and try to break it:
- Describe water pouring through a ceiling. Does it escalate, and does it mention the stop tap?
- Ask what a new boiler costs. Does it hold the line or invent a figure?
- Say you can smell gas. Does it direct you to the gas emergency line?
- Describe something genuinely ambiguous — “there’s a bit of damp under the sink” — and see whether it asks sensible follow-ups or files it flat.
- Ask for a job well outside your area.
- Then read the notification. If the address is wrong or the summary is vague, the call was wasted no matter how pleasant it sounded.
The honest summary
An AI receptionist does not make you a better plumber, and it does not solve a shortage of work. What it does is make sure that the people already trying to hand you work actually reach someone — at 2am, on a Sunday, while you are under a sink with both hands full.
For a plumbing business, more of that value sits outside business hours than inside them. Whether it is worth doing comes down to a number you can check today: how many calls you are currently missing.
Common questions
Can it handle a burst pipe call at 2am?
Yes, and this is where most of the value sits. It answers immediately, tells the caller to shut off the stop tap, captures the address and the severity, and escalates to you by text or a live transfer within seconds. The alternative at 2am is voicemail, and nobody with water coming through a ceiling leaves a voicemail — they call the next plumber.
Will it give a price for a blocked drain or a boiler swap?
It can quote a fixed callout fee or a published diagnostic rate, because those are real numbers. It should refuse to price a boiler swap or anything needing eyes on the job, and explain that pricing follows an assessment. An agent that guesses at those numbers creates an argument on site later.
How does it know whether a job is urgent?
It applies criteria you define — uncontrolled water, no water at all, sewage backup, no heat in winter, a leak reaching electrics. Everything else is routine. The important behaviour is that ambiguous descriptions escalate rather than being filed as routine, because the cost of a wrongly-escalated call is a phone buzz and the cost of a wrongly-ignored one is a flooded house.
What if I already have someone answering the phone?
Then it works as overflow and after-hours cover rather than a replacement — it picks up when the line is busy, at night and at weekends. Plenty of plumbing businesses run it that way and never point their daytime calls at it at all.
What makes the best AI receptionist for a plumber in Canada?
Judge it on behaviour, not features. Every caller should hear a recording notice in the greeting, and the agent should say it is an AI whenever a caller asks — US businesses disclose that up front. You keep the number on the van and the website by forwarding it, so nothing printed changes, and the same setup works for businesses in Canada and the US. Beyond that, three things decide it: how the escalation for a burst pipe or a no-water call is set up, whether it books the job or takes a message for a callback, and how it holds up on a live test call you make yourself.