Skip to content

Answering & missed calls

Do customers hang up on an AI receptionist? What callers react to and how to test it

By ApexScale AI · Published

The short answer

Some do, and some callers simply prefer a person. We found no published hang-up rate for AI receptionists in home service, so the only number to trust is your own call log. What tends to decide it is whether the caller gets what they called for: honest disclosure, a path to a person when it matters, and a callback promise you keep. See how the AI receptionist workflow sets those rules up.

The worry behind this question is fair: if a caller hears software, will they leave without leaving a message, and then call your competitor? The honest answer is that nobody has published a hang-up rate for AI receptionists answering plumbers, HVAC companies or roofers, so any confident percentage you read is a guess.

What you can do is learn what makes callers stay or leave, set rules that give them a way forward, and measure it on your own calls. This post covers all three, and says plainly when a person on the phone is the better choice.

Do customers actually hang up on AI receptionists?

Some do, and the only survey figure we found says many callers would rather reach a person. That is a stated preference, not a measured hang-up rate, and no published study tells you how many callers leave an AI receptionist for a home-service business. Your own call log is the number that counts.

The survey is from ServiceForge, a company that sells live answering services. Its website publishes a figure of 87% for callers who would rather talk to a real person than AI when calling a local service, attributed to a OnePoll survey. Its downloadable report, which we read on October 20, 2026, gives the question as asking people contacting a local service such as a plumber, electrician or roofing company whether they would rather speak to a real person or AI, and shows 87% real person, 6% AI and 7% no preference, from 6,000 respondents. The website dates the survey May 2026 and the report cover says April 2026.

Read it as a vendor survey from a seller of the alternative. A preference question also does not measure what a caller does at 10 p.m. with water coming through the ceiling. It is a reason to design for callers who want a person, not proof of how many will hang up.

What do callers react to on a call with an AI receptionist?

Based on how phone systems typically fail, not on a measured ranking, callers tend to react to five things: not knowing it is an AI, long menus, no way to reach a person, robotic repeated questions, and promises that are not kept. Each one is fixable in the setup.

None of these are unique to AI. A human answering service with a long script, a voicemail maze or an agent who cannot help produces the same reaction. What AI changes is that the failure repeats identically on every call until someone fixes it.

  • Being misled. A caller who finds out later that they were talking to software may feel tricked. Margot, ApexScale's AI receptionist for contractors, says she is an AI if a caller asks, and callers are told the call may be recorded.
  • Long menus. Callers who have to press numbers before anyone helps them are the classic ones to hang up. Test how many seconds pass before the first useful question.
  • No path to a person. When the problem is urgent, a dead end is what drives someone to call the next company. This is what a live-transfer rule is for.
  • Robotic loops. The same question asked three times, or "I didn't catch that" on repeat. Listen for it in test calls, and in real ones if your setup keeps recordings.
  • Broken promises. A callback promised in an hour that arrives tomorrow. Promise less and keep it.

How does a live-transfer rule change what callers do?

A live-transfer rule gives urgent callers a path to a person, which removes the main reason to hang up. Write down which calls ring you, what the caller is told, and what happens if you do not pick up. Callers with routine requests do not need the interruption.

Margot live-transfers urgent calls to you under rules you set, such as active water, no heat with a vulnerable person at home, or an existing customer with a visit today. Everything else is texted to you as a lead. Callers should be told in plain words what is happening, and if you cannot answer, the call should return to the receptionist to take details rather than dropping into your personal voicemail. Confirm exactly how that works in discovery and test it. Our live transfer or callback guide walks through writing those rules and what to ask a provider about unanswered transfers.

How does a clear callback promise change it?

A specific, believable callback promise gives non-urgent callers a reason to stay on the line and leave their details, because they know what happens next. Name a time frame you can keep and make sure someone owns the callback queue. A vague "someone will call you" invites the next search.

Margot texts you every lead as soon as the call ends, with the details you would otherwise ask for yourself, so your callback starts with the job, address and urgency instead of questions. Margot does not text the caller back. The promise to the caller is yours to keep, so set it to something your schedule can meet, such as first thing next morning for a routine request.

A callback is also where a human voice returns to the conversation. For many routine jobs, the person who calls back is what the customer remembers, not who answered first.

When does a live answering service fit better than an AI receptionist?

A live answering service fits better when many of your callers want or need a person from the first word: emotional or high-stakes calls, customers who dislike automated systems, callers who speak other languages, or work where every call needs judgment. An AI receptionist fits routine, predictable calls and overflow.

Many businesses use both: people for business hours or for callers who ask for one, AI for nights, overflow and simple requests. Our AI receptionist vs. answering service comparison lists public pricing and what to compare.

  • Callers who want a person. If your customers skew toward people who dislike talking to software, a person may keep more calls.
  • Sensitive or complex calls. After a storm, some homeowners want a person while they sort out damage and insurance. An AI can take details, but it does not discuss coverage.
  • Other languages. Margot answers in English only today.
  • Judgment calls. If most of your calls are unusual and need a person to decide, pay for one.

How can I test whether callers hang up on my AI receptionist?

Run practice calls before launch, then compare real calls that ended with a captured name and number against those that ended early. Add one call that asks for a person and one that does not answer a transfer. Your call log, not a survey, tells you how your callers behave.

To hear what your callers would hear, hear Margot answer as your business. Our evaluation guide turns the same idea into a longer pilot checklist.

  1. Before launch, make a handful of test calls with fictional scenarios: an emergency, a price question, an existing customer, a caller who says "can I talk to a person?", and a transfer you do not answer.
  2. Time how long it takes to reach the first useful question, and note any menu, loop or confusing wording.
  3. In the first weeks, count calls answered, calls that ended with name, number and the job captured, and calls that ended before that.
  4. If your setup keeps recordings or transcripts, listen to the early endings and look for the point where callers leave. Confirm what you can access during discovery.
  5. Compare against your baseline. Your missed-call audit shows how many calls went unanswered before, and how many would have been voicemail.
  6. Change one rule at a time, such as a transfer condition or the callback time, and compare again.

Frequently asked questions

Do callers know they are talking to an AI?

Margot answers in your business name and says she is an AI if a caller asks. Callers are told the call may be recorded. Listen to a test call to judge whether the greeting and the first question feel honest and quick.

What if a caller asks for a real person?

There should be a defined path: a live transfer if your rules say the call is worth interrupting you, or a callback with a time frame. Test this with a practice call before launch.

Is the 87% survey figure reliable?

It is a vendor survey from a company that sells live answering, with 6,000 respondents according to its report. It measures a stated preference, not hang-ups, and we treat it as direction rather than proof.

What about callers who do not speak English?

Margot answers in English only today. If a large share of your callers speak another language, a bilingual human service may serve them better.

Every lead lands on your phone

When Margot answers a call you can't, you get a text like this the moment it ends. Hear her answer as your business, or see your own numbers first with a free missed-call audit.

ApexScale AI9:18 PM
New lead
Chris P. · No hot water · asked for a callback tomorrow
Example text. Names and details are illustrative.

Sources

  1. Real People Answering Service - No Bots, No Call Centers, ServiceForge (vendor of live answering services; survey attributed to OnePoll, May 2026). Accessed October 20, 2026.
  2. Keep Service Human: What Real People Think About AI in Customer Service (survey report), ServiceForge (vendor survey, OnePoll, April 2026, 6,000 respondents per the report). Accessed October 20, 2026.

Third-party figures describe their publishers' own products, samples, or surveys. They are context, not ApexScale results.

Full refund if you cancel within 14 days of Margot answering your first call.