Do customers like AI answering services? What homeowners say
Homeowners accept an AI answering the phone if it's fast, says it's an AI, gets the job right and a human calls back in minutes. Here's what makes them hang up.
Customers don't mind an AI answering the phone, on four conditions. It picks up fast. It says it's an AI. It gets the job details right. And a human calls back within minutes. What they hate is hold music, voicemail, and a robot pretending to be a person. Get the four right and most homeowners shrug and get on with their day. Miss one and they've already called the next plumber on the list.
What homeowners actually say
One industry-specific example comes from Housecall Pro; it describes homeowners, not all customer audiences. They surveyed 1,040 US homeowners on 23 October 2025, all household decision-makers, and 53% said they were comfortable with AI handling the first contact. The same report says nearly every homeowner considers speed when choosing who to hire.
Read that the other way. Almost half aren't comfortable yet. That's not a reason to keep letting calls go to voicemail. It's a reason to build the thing so the uncomfortable half never has to spend more than a minute with it.
The wider consumer data is blunter. SurveyMonkey asked 2,017 US adults in December 2025 and found 79% strongly prefer talking to a human, 89% think a company should always offer a human option, and only 15% said the handoff from an AI to a person had ever gone smoothly for them. In the same survey, 14% said they'd lose trust in a business whose AI didn't make clear that it's an AI.
Those two surveys don't disagree. People want a human. They'll accept an AI for the first thirty seconds if a human follows. The number that should worry every vendor in this space is the handoff one. Most AI phone systems are sold on how well the robot talks. Almost nobody measures whether a person ever calls back.
Here's what I can and can't tell you from my own work. I have one plumbing company signed. Their system goes live in October 2026. The script was written from real call transcripts, not from a prompt I made up. I don't have my own customer-satisfaction data yet, and I won't pretend to. The first 30 days will go up on /built, good or bad.
The hang-up test: four things that lose the call, four that keep it
I call this the hang-up test. If a caller would hang up on a human doing this, they'll hang up on an AI doing it faster.
The four that lose the call:
- Slow. A two-second silence before it speaks, or a "hmm, let me think" after every answer. On the phone, silence sounds like a fault.
- Pretends to be human. A fake name, fake breathing, a fake "oh no, that sounds awful". The caller works it out in ten seconds and now everything else it says is suspect.
- Can't change gear for "water everywhere". The caller says the ceiling is coming down and the AI asks for the address in the same cheerful voice, then wants to know how they heard about you.
- Nobody calls back. The AI was polite, the job was captured, and then it sat in a spreadsheet nobody opened until Monday.
The four that keep it:
- Answers on the first ring. Faster than you can. That's the one thing it does better than a human, so let it.
- Says it's an AI in the first sentence. Not buried, not apologetic. Just a fact.
- Gets the job right and reads it back. What's wrong, where you are, how urgent. Three things. Then it repeats them so the caller hears they've been understood.
- A human calls back in minutes. Not "your enquiry has been received". A real plumber, with the details in front of them, while the caller is still standing in the kitchen.
Notice what isn't on either list. Nobody hangs up because the AI didn't book them a slot. They hang up because it wasted their time or lied to them.
The handoff is the product
Here's the honest version of what mine does. It answers on the first ring. It says it's an AI assistant. It finds out what the job is, where it is, and how urgent. It reads that back. Then it says someone will call shortly, sends the details to you, and gets out of the way. You call back in a few minutes and close it yourself, from your head, the way you always have.
It does not book. It does not promise a time. It does not say "someone will be there between two and four". Booking is a promise, and promises come from the person who has to keep them. A wrong booking costs you more than a missed call. You've driven across town for a job that was never real, or you've told a family "tomorrow morning" for a job that needed someone tonight. The AI has no idea what's on your van, what's on your board, or that the job before it just turned into a full afternoon. You do.
So why does this beat voicemail? Because nobody leaves a voicemail with water coming through the ceiling. They hang up and call the next name. If they do leave one, it's "hi, call me back, it's about the boiler", with no address, no callback number you can hear over the kids, and no sense of how bad it is. The AI gets all of that in under a minute, and the caller has been told, in plain words, that something is about to happen.
And why does it beat you answering with one arm behind a cistern? Because that call is bad for both of you. You're distracted, they're rushed, and you promise a time from memory that you'll forget by the time you've dried your hands. The AI takes the pressure off the first call. The second call, the one you make back, is calm. That's the one where the job actually gets won.
The caller's experience is two short calls instead of one bad one. Imagine it from their side. They call at half past nine. Something picks up straight away, says it's an AI, asks four questions, tells them a plumber will call shortly. Eight minutes later, a plumber calls. That's a better night than any voicemail has ever given anyone. More detail on the plumbing setup is on /ai-for-plumbers.
Say it's an AI in the first sentence
The greeting on mine goes roughly like this: "Hi, you've reached [your company]. I'm the AI assistant that takes calls when the team is on a job. Tell me what's going on and someone will call you back shortly."
Three reasons it's phrased like that. First, the customer finds out anyway. A caller who discovers it halfway through feels tricked, and the trick colours everything the human says afterwards. Second, it sets the expectation for the whole call. Nobody asks a disclosed AI for a firm price, and nobody blames it for not knowing your diary. Third, there are rules.
In the EU, Article 50 of the AI Act has applied since 2 August 2026. It requires AI systems that interact directly with people to be built so that the person is told they're dealing with an AI, at the latest at the first interaction. The Commission's own FAQ confirms the date and names chatbots and AI agents as the kind of system it means. The UK isn't in the EU, but plenty of vendors build one script for both.
In the US it's state by state. Utah's S.B. 226, in force since 7 May 2025, says a business using generative AI in a consumer transaction must admit it's AI if the customer asks, and it gives a safe harbor to anyone whose AI "clearly and conspicuously" says at the outset that it's an AI assistant. California's older bot law, Business and Professions Code 17941, covers online bots in commercial interactions and exempts anyone who discloses. Other states are moving. I'm a plumber, not a lawyer, so check with your own advisor for where you operate.
Here's the practical point. Disclosure costs you nothing. The caller who wanted a human still gets one, in minutes. The only thing you lose is the option to fake it, and faking it was never going to survive a second call.
Call your own line tonight
Before you buy anything from me or anyone else, run this test. It takes one evening and costs nothing.
Call your own business number at half past nine tonight, from a phone the system doesn't know. Be a customer. Say the words "water everywhere". Then start a stopwatch and time how long it takes before a human being calls you back.
That number is the whole business. Not the AI's voice, not the dashboard, not the integration list. If the answer is "never", that's your answer, and it's the same answer your customers have been getting. If it's forty minutes, you've found out why your after-hours calls don't turn into jobs. If it's eight minutes, don't change anything.
While you're on that call, check three more things. Did it tell you it's an AI before you had to ask? Did it change gear when you said "water everywhere", or did it plod through the script? And when you asked "are you a real person", did it tell the truth?
If you want the missed-call maths for your own numbers, the calculator will do it in five minutes with your own figures. And if you decide you want to talk it through, book a growth call. Thirty minutes, no obligation, and I'll tell you if I think you shouldn't buy it. I say that more often than you'd expect. If your phone rings twice a day and you answer both, keep your money.
Questions people ask
Do customers hang up on AI receptionists?
Some do, and it's almost always for one of four reasons: it was slow, it pretended to be human, it couldn't change gear when the caller said water was everywhere, or nobody ever called back. Fix those four and the hang-ups mostly stop. Housecall Pro's 2025 homeowner survey found a majority already comfortable with AI handling the first contact.
Should the AI say it's an AI?
Yes, in the first sentence, as a plain fact. Callers find out anyway, and the ones who find out late feel tricked. It also sets expectations: nobody asks a disclosed AI for a firm price. The EU AI Act requires it from August 2026, and Utah gives a safe harbor to businesses whose AI says so at the outset. Check with your own advisor for your area.
What should happen after the AI takes the call?
A human should call the customer back within minutes, with the job details already in front of them. The AI's job is to catch the call, find out what's wrong, where and how urgent, read it back, and tell the caller someone will call shortly. It should not book a slot or promise a time. That's yours to do, from your own diary.
Is an AI receptionist better than voicemail?
For urgent trades work, yes. Nobody leaves a voicemail with water coming through the ceiling; they call the next plumber. An AI that answers on the first ring gets the address, the problem and the urgency, and tells the caller a human is coming back to them. Voicemail gets 'call me back' and no details, if it gets anything at all.
Can an AI receptionist handle emergencies?
Partly. A good one recognises urgency words like flood, burst, no heating or smell of gas, skips the routine questions, gives the caller the emergency number you've set, and flags you straight away. What it can't do is dispatch. It has no idea where your van is or what's on your board. The human callback is what handles the emergency.
Sources
- 1.Housecall Pro: Home service customer service report (2025) · 1,040 US homeowners, fielded 23 October 2025; 53% comfortable with AI handling the first contact
- 2.SurveyMonkey: Customer service statistics, humans vs AI (December 2025) · 2,017 US adults; 79% prefer a human, 89% want a human option, 15% had a seamless handoff, 14% lose trust without disclosure
- 3.EU AI Act, Article 50: Transparency obligations for providers and deployers · Applies from 2 August 2026; people must be told they are interacting with an AI system at the latest at first interaction
- 4.European Commission: FAQ on transparency obligations under Article 50 · Confirms the 2 August 2026 date; names chatbots and AI agents as systems covered
- 5.Utah S.B. 226 (2025): Artificial Intelligence Consumer Protection Amendments, enrolled copy · 13-75-103 disclosure if the consumer asks; 13-75-104 safe harbor for clear disclosure at the outset; effective 7 May 2025
- 6.California Business and Professions Code, section 17941 · Online bot disclosure in commercial interactions; a person who discloses is not liable

Written by Ivar André Knutsen
I build and run AI systems, internal tools and workflow automation. You work directly with me from the first conversation through implementation and support. About Ivar
Want this looked at in your business?
We look at where you want the business to go, what is slowing you down and where AI could make a useful difference. You get a clear recommendation: a tool to try, a focused automation, a broader system or a closer look at the process. Any build is scoped and quoted before work starts. Free, no obligation.
Single automations are quoted on the call: a setup fee plus a monthly retainer to run them. Full systems start at $4,500, fixed scope, fixed price.
Book a free 30-minute call