What customers write about AI in your front office, and what leading operators say is working in 2026.
of reviews that commented on a company's AI were negative about it
Genuine praisepraised a company's AI, all for an after-hours booking a person honored
A person on the phonewere five-star reviews praising a company for not using AI
National moodof US adults trust AI only a little or not at all (Morning Consult)
If you run a home services business with 100+ employees, you've already made an AI decision, or you're about to. Intake bots greet inbound calls. AI receptionists book appointments after hours. Automated texts handle reminders and review requests. Most of it happens without the customer being told, and most of it is invisible until something goes wrong.
We wanted to know what customers actually think about all of it, so we went to the most honest data source available: Google reviews. We read 10,938 reviews of HVAC, pest control, plumbing, electrical and roofing companies and pulled out every one that mentioned AI, chatbots, virtual assistants or automated phone systems.
Customers mention AI in two ways. Some name it outright: the tool, the chatbot, the receptionist, the brand. Others signal it by asking for a human, praising a real person, or complaining about an answering system. Both carry the same emotional weight.
Each one is a fixable architecture problem, not an AI problem.
"The AI booked me for a time slot they didn't have, and I didn't have an appointment. An answering machine would have been better customer service."HVAC customer · 1-star review
The most damaging reviews, where customers cancel service, switch providers or warn others, almost all share one feature. The customer was in an emergency, and the AI controlled the path to a person.
Ice storms. No water for five days. A $16,000 install that left tenants without heat. A multi-day plumbing failure where the only contact channel was an AI assistant. These are the moments that decide whether a customer becomes a lifetime account or a one-star review with a screenshot.
"We spent $16k and have no heat or hot water. My tenants are now at risk of leaving. There is no emergency line to call because it is AI. We are desperate."From the dataset · HVAC customer
The positive side of the ledger is starker than the negative. Only one kind of review counted as positive toward a company's AI, and it is a 6% slice.
"Huge fan of their AI phone reception active 24/7 for instant booking. Was able to call at 2 a.m. and the AI confirmed my appointment for 7 a.m."HVAC customer · 5-star review
The recut sharpens the playbook. The 6% who reward deployed AI are describing one specific thing done right. The 13% rewarding its absence are a reason to keep the human-first experience intact even as competitors automate it away.
Every cost case for AI in the front office leads with the same number. An AI receptionist costs a few dollars per resolution. A local CSR earning $20–$25 an hour costs $35–$45 fully loaded. On paper the gap looks decisive. In practice it almost never is, because of three costs almost no one runs the numbers on.
"Companies cutting the most workers showed nearly identical financial returns to those cutting the least. Workforce reductions create budget room. They do not create return."Gartner · May 2026 survey of 350 enterprises deploying AI
Two themes show up in every operator conversation we've followed. The companies winning in 2026 aren't the ones with the most AI or the least. They're the ones who settled two questions the rest of the industry is still arguing about.
The operators winning with AI aren't running a labor-replacement strategy. They're running a friction-removal strategy. AI earns its place where the work is repetitive, low-judgment and high-volume, and it frees people for the moments that drive revenue.
"AI does not replace people. It substitutes for the requirement that people perform dumb tasks. You don't need it to turn a wrench. You need to use it to make your humans superhuman."Tersh Blissett, CEO, Service Emperor HVAC · Contracting Business, July 2025
Chad Peterman runs Peterman Brothers, a $120M+ HVAC, plumbing and electrical company across Indiana and Ohio. His take is operational economics. Demand spikes with weather, and you can't staff peak-season call volume without burning money in the slow months. AI helps with that math, and it filters the noise.
"50 to 70% of the calls that come into our contact center are not actual leads. That's where AI becomes impactful. It scrubs the junk off our CSR's plate so they can focus on the interactions that customers really value."Chad Peterman, President & CEO, Peterman Brothers · Hatch AI case study, 2025
Every operator we listened to lands on the same point. The failures customers write reviews about aren't AI failures. They're architecture failures. The companies getting this right decided in advance which work AI owns, which work people own, and how the hand-off happens.
"This isn't about tech vs. people — it's about making sure the right person is in the seat. And sometimes, that person is AI."Tim Brown, Founder, Hook Agency · Hook Agency blog, 2025
"AI is just another tool in the toolbox — powerful when used with purpose, dangerous when used without systems."Al Levi, Founder, The 7-Power Contractor · The 7-Power Contractor blog, 2025
Best Choice Roofing, the fourth-largest residential roofing company in the US, spent the past year overhauling its tech stack. CEO Bryce Barnett reached the same conclusion the review data did: customer pushback on AI is real, and it's louder than most CEOs realize.
He starts every AI decision with two questions: can AI replace this job, or can it make a person more efficient? "It doesn't have to be 'or,'" he says. "It's always 'and.'"
"More and more people are realizing that AI is everywhere, and then they're also saying, 'I don't want that.' I actually want to deal with people."Bryce Barnett, CEO, Best Choice Roofing
Best Choice writes 600–700 insurance contingencies a year: agreements to work with homeowners and their insurers on a claim. Seven internal estimators couldn't keep up. Fewer than 25% of contingencies had an estimate attached, so the team was effectively choosing which homeowners to help. An early custom AI agent helped, but it couldn't carry the volume.
"We couldn't deploy enough AI because there wasn't enough custom AI to be able to go build this yet."Bryce Barnett, CEO, Best Choice Roofing
Instead of hiring local estimators at about $75,000 a year, Best Choice added 12–15 Hire Bloom team members to the estimating team and gave them software to write estimates faster and more accurately. The AI stayed in place, and the people worked alongside it. The partnership has since grown to 22 Hire Bloom team members across the business.
Savings are estimated at $21,734 per team member, based on Hire Bloom's average client savings, and exclude the value of estimates that previously went unwritten.
Put the review data, the cost math and the operator commentary side by side and there's no AI-versus-humans war to win or lose. The work-allocation matrix splits front-office work on two questions: how routine is the task, and how much is at stake for the customer if it goes wrong?
Home services has a customer-base problem most consumer industries don't. Nearly two-thirds of Americans now distrust AI, and the distrust is concentrated in exactly the people who buy home services.
The homeowner most likely to call about a roof leak, a furnace or a pest problem is also the homeowner most likely to hang up on an AI receptionist. Industry trust in AI fell year over year for the first time in the May 2026 reading. Companies treating AI as a quiet infrastructure layer behind a human-led experience are aligned with where their customers are. Companies putting AI on the first call are not.
Source: Morning Consult, The AI Trust Report, May 2026.
"Industry trust in AI peaked in October 2025 and has softened in subsequent readings. The companies growing fastest are not the ones absorbing that headwind. They're the ones routing around it."Morning Consult · The AI Trust Report, May 2026
Each comes from the review data and from how operators staff their front office. Each is a moment AI is structurally bad at and a trained person is structurally good at.
The 10,938 reviews come from three review sets.
A first pass searched every review for eight categories of language. A second pass read every match by hand to assign categories, score sentiment and remove false positives. Four duplicate reviews across the sets were counted once, and 18 false positives were removed, most often "ai" typed for "I."
Each review was scored on what the customer said about the AI, not on the review's star rating. A five-star review can carry negative AI sentiment when the customer praised the technician but criticized the AI receptionist. Eleven reviews that mentioned AI only as the customer's own research tool ("ChatGPT recommended them") were excluded, since the company's AI never entered the experience. Five-star reviews praising a company for not using AI were scored as their own category rather than as positive toward AI.
Free, no form. The PDF carries the same findings, figures, and methodology as this page.
Hire Bloom is the staffing partner for remote support roles. We help home services companies build their support teams — customer service, scheduling, quoting, and back office — with US-educated team members who embed directly in your team and stay: 70% are still on the job a year later. Flat, predictable pricing starting at $13/hour. Not a BPO. Not AI. Your team.
We publish home services research to help operators think clearly about the decisions in this report: where AI fits, where people do, and how to grow the back office without losing the customer experience that earns the next job. If the patterns here match what you're seeing in your own business, we're happy to talk them through — you'll find a way to reach us below.
Tell us where you're at, and we'll show you what's possible.
Find the right talent for your open roles
Get a custom plan with flat, predictable pricing
Start interviewing in as little as 2–3 days
