Do customers mind talking to AI on the phone

The objection I hear most from plumbers goes like this. My customers are older, they want a person, I am not putting a robot on my phone. Fair enough. So what happens right now when you are under a house and the phone rings? It goes to voicemail. And how many of those callers leave a message? Most owners have never counted, and the ones who do are usually surprised how low it is. So the choice was never a robot or a person. It was a robot or a beep. That is the honest version of the question in the title, and the answer is more interesting than yes or no.

01The comparison most owners get wrong

When owners picture a customer talking to an AI, they compare it with their best day. You, or your office manager, picking up on the second ring, knowing the customer by name, remembering the water heater you put in three years ago. Of course the AI loses that comparison. Nothing beats a good person who knows the account.

But that is not the call anyone is talking about replacing. The call in question is the one that already reaches nobody. You are on a roof. Your office manager is on the other line. It is 7:40 at night. The phone rings four times and the caller hears your greeting and a beep.

So the fair test is not “AI versus my best receptionist.” It is “AI versus what that caller gets today.” For most small shops, what they get today on a missed call is voicemail, and a lot of people hang up on voicemail without saying a word. They call the next name in the search results instead.

The research on how fast businesses respond is not flattering, and it is worth knowing even though it measured web inquiries rather than phone calls. Harvard Business Review audited 2,241 US companies and timed their responses. The average first reply took 42 hours. 23% never replied at all. A caller who gets a beep at your business has been trained by experiences like that to assume nobody is coming back.

02What customers say when someone asks them

The honest answer is that nobody has published a big, clean survey of plumbing or HVAC customers on this exact question. I would rather tell you that than dress up a number from somewhere else.

The closest real data I have comes from a different industry. Zenoti, a software company that serves med spas, reported that around 71% of those clients are comfortable with an AI handling some of their interactions. That is a spa client booking a facial, not a homeowner with a leaking pipe, so do not carry the number over directly. What it tells you is narrower and still useful. A clear majority of ordinary consumers are no longer shocked by the idea, and the word “some” in that finding matters. People are fine with AI handling parts of an interaction. They are less fine with it handling all of it.

Listen to how people describe a bad automated call and the complaint is rarely “I talked to an AI.” It is about a specific failure. It did not understand me. It kept asking the same thing. It would not let me reach a person. It told me something that turned out to be wrong. Every one of those is a design problem, and every one of them can be tested before you put anything on your line.

03What people hate is the phone tree, not the voice

Think about the calls you personally dread. Press one for billing. Press two for support. Say “representative.” I didn’t catch that. Please say “representative.” You have been through that, and so have your customers, and it shaped how everyone feels about talking to a machine.

Most of what people remember as a bad robot call was a menu. It made you do the work. It forced you through its categories instead of listening to yours, and it hid the exit.

A decent AI answering system does the opposite. The caller just talks. “Yeah, hi, my AC is blowing warm air and I’ve got a baby in the house.” It should take that sentence as it comes, ask one or two follow-up questions a normal person would ask (name, address, best number, when it started), and then tell the caller what happens next.

There are three things that make callers angry, and you should listen for all three when you try any system:

1. It pretends to be a person, and the caller figures it out halfway through.
2. It traps them. There is no way to reach a human, even when the situation clearly needs one.
3. It makes a promise the business cannot keep. It books a time, quotes a price, or says someone will be there in an hour, and then reality arrives.

Get those three right and the voice itself fades into the background for most callers. Get any of them wrong and no amount of natural sounding speech saves it.

04Tell them it is an AI, every time they ask

This is the part where I get opinionated. If a caller asks “am I talking to a real person,” the only acceptable answer is the true one.

Part of that is the law moving in one direction. In February 2024 the FCC made clear that calls using AI generated or artificial voices fall under the TCPA, and outbound calls like that generally need the called party’s prior express consent. That ruling is about calls a business places. A customer dialing your number is a different situation. But the direction is plain. Regulators do not like machines passing themselves off as people, and neither do customers.

The bigger reason is trust. A homeowner who finds out mid-call that the friendly voice was a machine does not think “how clever.” They think “what else is this company hiding.” That is a bad feeling to plant in someone who is about to let your tech into their house.

In practice, disclosure costs you very little. Most callers do not ask. The ones who do usually just want to know whether they can get a person, and the answer to that should be yes, with a clear path. A caller who hears “I’m an automated assistant for Martinez Plumbing, I can take down the details and get this to the team right now” gets on with it. They called because something is broken. They want it fixed more than they want a philosophical conversation.

Recording deserves the same honesty. If the system records calls, and it should because that is how you check its work, it has to say so. A handful of states require every party on a call to consent to recording, so the announcement is not optional.

05Where customers are right to push back

I would be selling you something if I said everyone is fine with this. Some callers are not, and some of them have a point.

Emergencies are the clearest case. A person who smells gas, sees sparks, or has water pouring through a ceiling does not want a pleasant conversation. They want a human who can act. A good system hears those words and hands the call to a real phone right away, without collecting an email address first. It should never try to diagnose what is wrong or tell someone whether they are safe. That is not a job for a machine, and a vendor who says otherwise is telling you something about how they think.

Pricing is another. Callers want a number. An AI that guesses a price over the phone sets a ceiling that your tech then has to break in the driveway. The only number it should ever say is one you decided in advance, like a trip charge. I wrote about how that goes wrong in the after-hours call you should not quote over the phone.

Then there are your regulars. Somebody who has used you for ten years and has your cell number probably should not be routed through anything. That is fine. The system handles the calls that would otherwise go nowhere. Your best customers can keep calling you the way they always have.

And yes, some people just do not like it. A few will hang up on any automated voice. You cannot win every caller. What you can do is compare that small group with the much larger group hanging up on your voicemail today.

06How to find out what your own customers think

You do not have to take my word or anybody’s survey. You can check this on your own line in about two weeks.

Start by calling your own business number from a phone that is not yours, at a time when nobody would pick up. Sit through the whole thing. Ask yourself honestly whether you would leave a message.

Then count. Pull last month’s call log and see how many calls went unanswered and how many of those left a voicemail. The method is in how to count the calls your business is missing. That number is your baseline, the share of missed callers who stuck around to talk to your current setup.

If you try an AI answering system, judge it on the same terms. Of the missed calls it picked up, how many callers stayed on and gave their details? How many asked for a person, and did they get one? How many hung up in the first ten seconds? Then do the part most owners skip and listen to the recordings. Ten calls is enough to hear whether people sound annoyed, relieved, or indifferent. You will know within a few minutes of listening.

If you want a list of things to check before paying anyone, I put one together in what to ask an AI phone vendor before you pay them anything. And for a fair look at where a human is still the better answer, read AI vs a human receptionist.

Frequently asked questions

Will older customers hang up on an AI?

Some will, and some already hang up on voicemail. Age matters less than how the call is built. A system that lets people talk normally, says plainly what it is, and gets them to a person when they ask for one is easier for an older caller than a phone menu. Listen to the recordings from your own older customers before you decide.

Should the AI say it is an AI at the start of the call?

It should never hide it, and it must answer truthfully when asked. Whether it announces itself in the first sentence is a judgment call. Many businesses use a greeting like “automated assistant for” plus the company name, which is honest without making a speech about it. What is not acceptable is a system built to pass as a human.

Can customers tell the difference anyway?

Often, yes, especially on a longer call. That is one more reason not to pretend. The goal is not to fool anybody. The goal is that a caller who would otherwise reach a beep gets their problem written down and sent to you while they are still on the line.

What if a customer insists on talking to a person?

Then they should be able to. A good setup either transfers the call to a human number or takes their details and tells them exactly when someone will call back. Trapping a caller who asked for a person is the fastest way to make someone angry at your business.

Will this change the number my customers call?

No. Your number stays the same. Calls you do not answer are forwarded to the system, and calls you do answer work exactly like they do now. The details are in what happens to your phone number when you add an answering system.

The bottom line

Most customers do not mind talking to an AI as much as owners fear. What they mind is being tricked, being trapped, or being told something that turns out to be false. On the calls that reach nobody today, an honest system that takes the details and gets them to you is a better experience than a beep, and your own recordings will tell you whether that holds for your customers.

These days we book appointments for roofing companies: we find homeowners looking for roof repair or replacement, answer their replies and book them into your system. If you run a roofing company, see how it works at rushdial.com/roofing.