Short answer: Almost never, based on one year of our own call data. Across 118,683 real calls to 469 service businesses using Marlie’s AI answering service (November 2025 to October 2026), only about 0.1% of non-spam calls showed any objection to talking to an AI, and on 37 calls (0.03%) the caller hung up because of it. About 1.2 to 1.5% of callers asked “Am I talking to a real person?”, and roughly 40% of them stayed and finished the call. On the other side of the ledger, 22.5% of calls came in outside business hours, 1 in 5 leads arrived after hours, and 25% of the jobs the AI booked came from after-hours calls. The real risks are elsewhere: callers who hang up silently, transfers nobody picks up, and complaint calls.
| What we measured (Nov 2025 to Oct 2026) | Result |
|---|---|
| Calls where the caller objected to the AI | About 105 calls, 0.1% of non-spam calls |
| Callers who hung up because it was an AI | 37 calls, 0.03% |
| Callers who asked “Am I talking to a real person?” | 1.2% to 1.5% of calls; about 40% of them finished the call |
| Callers who asked for a person | 16% to 18% of non-spam calls |
| “Human, please” requests transferred, where live transfer is set up | 85% |
| Calls outside business hours | 22.5% |
| Leads that came in after hours | 1 in 5 (4,226 of 21,719) |
| AI bookings that came from after-hours calls | 25% (435 of 1,730) |
| Callers who called the same business again within 90 days | 21% |
Data: every Marlie customer business, including ones that later left. 118,683 real calls at 469 businesses; internal and test calls removed. Full method below.
The fear: “our customers like the personal interaction”
The most common worry we hear from owners on sales and onboarding calls is not price. It is that callers will hate it.
Aaron Blattner of West Shore Mechanical put it plainly: “our customers like the personal interaction as opposed to just AI.” Tammy, who runs the EyeCandy Cape salon, asked what happens when callers “recognize, oh this is a robot and they just keep pressing zero zero zero,” and worried about “older clients that are not in tune to technology.”
These are fair questions, and most articles answer them with opinion. We had a year of calls, so we counted.
How many callers object to talking to an AI?
About 105 calls out of all non-spam calls in the year showed an objection to the AI. That is 0.1%.
On 37 of those calls (0.03% of the total), the call summary says the caller hung up because they were talking to an AI. Most of the other objectors were transferred to a person or kept talking and finished what they called about.
Put another way: for a business that gets 58 real calls a month (the typical Marlie business in October 2026), an objection that ends the call would happen about once every four to five years.
“Am I talking to a real person?” What happens next
Between 1.2% and 1.5% of calls included a caller asking whether they were speaking to a real person. Marlie answers honestly: it says it is the business’s AI receptionist.
About 40% of the callers who asked stayed on the line and finished the call: they booked, left their details or got their answer. The rest asked for a person, were transferred or ended the call.
The question also fades with familiarity. Callers asked it on 1.5% of first contacts, but only 0.4% by their third call to the same business. Once a caller knows what to expect, they mostly just get on with it.
How often do callers ask for a human, and do they get one?
Between 16% and 18% of non-spam calls included a request to speak to a person. That is more than 100 times as common as objecting to the AI. Asking for a person is a normal part of a business call, not a verdict on the AI.
What happens next depends on how the owner set it up:
- Where live transfer is set up, 63% of callers who asked for a person were transferred, and 85% of explicit “human, please” requests were. Marlie can send a call to the owner’s cell, a specific staff member, or a different person for each call reason, using warm transfer.
- Where nobody is free, or live transfer is off, the AI takes a message and sends it to the right person by text or email.
One honest catch: a transfer only helps if someone picks up. In October 2026, 64% of transfers were answered. The other 36% were not picked up, which is the same busy-owner problem the AI was covering in the first place. If you turn transfers on, decide who answers them and when.
Caller mood: positive, neutral and negative calls
From June 29, 2026, every call got a sentiment score. Across 38,546 scored calls (June to October 2026):
| Calls | Share negative |
|---|---|
| All scored calls (69% neutral, 13% positive) | 18% |
| New-lead calls | 13% |
| Complaint calls | 79% |
| All calls, July 2026 | 19% |
| All calls, October 2026 | 16% |
Most calls are neutral, which is what you want from a phone call about a leaking pipe or a haircut. New-lead calls, the ones that make you money, were negative only 13% of the time. Complaints were negative 79% of the time, but those callers were already upset when they dialed.
Negative calls also fell over time, from 19% in July to 16% in October, as owners tuned their agents’ answers, voices and rules.
Do callers come back after reaching an AI?
Yes, at the same rate all year. 18% of callers called the same business again within 30 days and 21% within 90 days. If callers were being driven away by the AI, repeat calls would drop as more of a business’s customers met it. They did not.
Does it sound like a robot? What owners change after the first test call
Owners tend to worry less about the AI itself than about how it sounds. Michael Cohen of White Swan Auto Spa in New York said: “I don’t want it to be like, uh, like a AI reception. I want to be like a more of a human being.” Derek Gibson of Dr Sparky LLC asked whether it gets “softer,” because “I’m in the South man. People are soft.” Larry Debenedictis of Fast Mobile Truck Repair listened to a test call and said it was “talking too fast a little bit, yeah, especially in the introduction.”
These are settings, not limits. In Marlie, owners pick from 69 voices filtered by gender, age and language, hear each one say their own greeting before it goes live, choose a personality (Friendly & Warm, Formal & Professional, Quick & Efficient, or one they write themselves), and set the speaking pace. Background noise stays off unless the owner wants a quiet office sound.
Most owners do change something. Of 155 live businesses, 99 picked a voice other than the default, 37 different voices are in use, and 109 wrote their own greeting. Friendly & Warm is by far the most popular personality (115 businesses).
What businesses got: after-hours calls, leads and bookings
The flip side of “will it hurt?” is “what would I miss without it?” Here is what the AI picked up that would otherwise have rung out or gone to voicemail.
After-hours calls are a fifth of the phone
At businesses with set hours, 22.5% of calls came in outside those hours (the median business: 22%). 12% of all calls came on weekends.
After-hours callers are real leads
The AI logged 31,982 lead calls from December 2025 to October 2026. 32% of them came on weekends, before 8 a.m. or after 6 p.m. At businesses with set hours, 1 in 5 leads came after hours (4,226 of 21,719, January to October 2026).
After-hours callers were not tire-kickers. At the same business, the median share of calls that were leads was 30% during business hours and 30% after hours (72 businesses, January to October 2026).
A quarter of bookings happen while you are closed
From April 24 to October 31, 2026, the AI booked 1,965 jobs at 44 businesses. At businesses with set hours, 25% of those bookings (435 of 1,730) came from after-hours calls. Those are jobs an owner wakes up to, instead of voicemails to return.
If most of your missed calls come at night and on weekends, an after-hours answering service is the first place an AI earns its keep. It also matters if you buy Google Local Services Ads: since October 2026, Google can charge you for missed calls during business hours.
A typical month
A typical live business got 58 real calls and 9 leads in October 2026. A business that books through the agent gets about 4 to 5 bookings a month from it.
Where AI answering does lose callers (honest limits)
The 0.1% objection rate is real, but it is not the whole story. These are the places we see callers slip away:
- Silent hang-ups. 20% of calls end with the caller saying nothing, after a median of 7 seconds. Many are robodialers, wrong numbers and pocket dials, but some are people who hang up when they hear an AI. 38% of silent callers called back within 7 days. Between 12% and 16% of first-time callers hung up this way and never talked to the agent later. That is the clearest “lost caller” number in the data, and the upper limit on callers the AI greeting may have turned away.
- Transfers nobody answers. Only 64% of transfers were picked up in October. A caller who asks for a person and then gets a second voicemail is worse off than one who got a good message-taker.
- Complaints. 79% of complaint calls were negative. An AI can log the complaint and get it to the right person fast, but it will not calm an angry customer the way the owner can. Route complaints to a person.
- Callers who don’t like tech. Owners like Tammy worry about older clients. We could not measure caller age, but this is one reason to keep a clear path to a human and a simple greeting.
- Businesses where every call needs the owner. If you have to see the job before giving a price, or you take no payment until you have spoken to the caller (as Larry does), the AI’s job is to collect details and pass them on, not to close.
What owners told us
Jason Gilbert of Gilbert Engineering USA, about one of his first real calls: “I saw one of the customers called in yesterday and actually talked to it and it was, it was good.”
Larry Debenedictis runs Fast Mobile Truck Repair, a one-man roadside truck repair business. He does all the work himself, so he often can’t pick up the phone. His view of AI answering is practical: “it’s better than not picking up the phone at all.”
Aaron Blattner, in the same conversation where he said his customers like the personal touch, explained that every after-hours call goes to his cell, and “it’s not uncommon to get 12 to 20 in a weekend. I would rather have AI answer that phone call.”
That is the pattern we see in the numbers too. Owners rarely choose AI over a person. They choose it over voicemail, a missed call or a 2 a.m. wake-up.
How to keep an AI answering service from hurting your business
Based on the data, and on what owners change after their first calls:
- Let the AI say it is an AI. Callers who ask get an honest answer, and 40% of them finish the call anyway.
- Give callers a way to a person. Set up transfers for the people and hours that can actually pick up, and a message for everything else.
- Pick the voice and pace yourself. Test call it. Most owners change the default.
- Send complaints to a human. Let the AI handle the logistics and you handle the feelings.
- Start where you are losing calls already. For most service businesses, that is after hours, weekends, and the hours you are on a job. Marlie can also act as an overflow answering service that picks up only the calls you can’t.
- Check the calls every week for the first month. Every Marlie call comes with a recording and transcript plus a summary, so you can see exactly what callers heard.
Marlie answers in 2 seconds and works with HVAC, plumbing, electrical, roofing, salons, property management and other home service businesses. Plans start at $49 a month, with a 14-day trial (see pricing).
Method and sample
- Who: every Marlie customer business from November 2025 to October 2026, including businesses that later cancelled.
- Caller-side numbers: 118,683 real calls at 469 businesses. Business-side numbers (hours, leads, bookings) use a slightly different test-call filter: 119,242 calls at 471 businesses. 80% of the calls came in May to October 2026, as the customer base grew.
- Removed: internal calls and test calls made by owners and our team. “Non-spam” excludes calls marked as spam.
- Objections, “real person?” questions and requests for a person were counted from call summaries, then re-counted independently with a second set of queries. Where the two counts differed, we give the range (for example, 1.2% to 1.5%).
- Sentiment has been scored on every call since June 29, 2026 (38,546 scored calls through October).
- After hours means outside each business’s own set hours; businesses without set hours are left out of the after-hours figures.
- Snapshot date: data pulled read-only from production on November 3, 2026. Customer businesses are not named; owners quoted gave their words on recorded calls with us.
For how Marlie compares with other tools, see our test of the best AI answering services, and for the human side of the choice, AI receptionist vs human receptionist.
FAQ
Will an AI answering service hurt my business?
In our data, rarely. Across 118,683 calls at 469 service businesses, 0.1% of non-spam calls showed an objection to the AI and 0.03% ended because of it. The bigger risks are silent hang-ups from first-time callers (12% to 16% never talked to the agent later), transfers nobody answers, and complaint calls, all of which you can reduce with setup.
Do customers hang up on AI receptionists?
A few do. 37 of 118,683 callers (0.03%) hung up after objecting to the AI. Another 20% of calls end in silence after about 7 seconds, but many of those are robodialers, wrong numbers and pocket dials, and 38% of silent callers called back within a week.
Do callers know they are talking to an AI?
Some ask. 1.2% to 1.5% of callers asked “Am I talking to a real person?” Marlie tells them it is the business’s AI receptionist, and about 40% of them carried on and finished the call. Returning callers ask far less: 0.4% by their third call.
Can callers still reach a real person?
Yes, if you set it up. 16% to 18% of callers asked for a person. Where live transfer is on, 85% of explicit “human, please” requests were transferred. When nobody is free, the AI takes a message and sends it to the right person.
Is an AI answering service worth it for a small business?
It depends on how many calls you miss. In our data, 22.5% of calls came after hours, 1 in 5 leads came after hours, and 25% of AI bookings came from after-hours calls. A typical business got 58 real calls and 9 leads a month. If you already answer every call yourself, the gain is small. If you miss calls on jobs, at night or on weekends, those are the calls the AI catches.
Does an AI receptionist sound robotic?
It can if you leave the defaults. Owners most often adjust the voice, pace and warmth after their first test call. 99 of 155 Marlie businesses picked a voice other than the default, and 109 wrote their own greeting.
Is AI answering better than voicemail?
For most callers, yes. A voicemail takes no details, books nothing and often gets no message at all. In this study the AI booked 1,965 jobs at 44 businesses and logged 31,982 lead calls. As one owner put it, “it’s better than not picking up the phone at all.”

