TL;DR
- Contractors are not quietly sceptical about AI answering services. In trade forums they are openly hostile, and much of the hostility is earned by bad products.
- The most useful evidence we found is an owner who tested one honestly: of seven calls, it helped with two, and callers hung up on the bot more often than they did with his answering service.
- The objection is not really about AI. It is about a caller with no heat in January getting a machine that cannot help and cannot transfer.
- Four conditions separate a system that books jobs from one that loses them: it answers instantly, it recognises an emergency, it hands off to a human on demand, and it never pretends to be a person.
- If you personally answer nearly every call, you do not need this. It earns its place when the phone rings while you are physically unable to pick up.
We sell an AI phone system, so treat everything here accordingly. We are also the ones who went and read what contractors say about products like ours when no vendor is in the room, and it is not flattering.
A roofer summarised the mood in four words: an AI assistant would lose more jobs than missed calls. On a thread where an owner asked how to stop missing calls, the top comment was someone predicting, with visible weariness, that the next reply would be a pitch for an AI phone bot. Another owner, writing a list of requirements for an answering service, put it in his specification: he did not want an AI assistant, because they suck.
Most vendors respond to this by talking louder about accuracy. We think the objection deserves to be taken at face value first, because a good deal of it is correct.
The most honest test we found
The single most useful piece of evidence was not a case study. It was a small business owner reporting on his own trial, without a product to sell.
Of his last seven calls, the bot had been able to help with two. It successfully screened one out and booked another. Beyond that, he said, numerous people simply hung up on it, and they hung up at a noticeably higher rate than they did with the human answering service he had been using.
Two out of seven is not a disaster and it is not a triumph. What makes the number worth sitting with is the comparison. He was not measuring the bot against voicemail. He was measuring it against the humans it replaced, and on that comparison it lost.
That is the number every vendor should be asked for and almost none will volunteer: what share of callers hang up? If a salesperson cannot tell you, they either have not measured it or would rather you did not know.
What the hostility is actually about
Read enough of these threads and the complaint is never really about the technology. It is about four specific failures, each of which is a choice somebody made.
- It cannot tell an emergency from a price shopper. A no-heat call in January and a request for a maintenance quote are not the same call. A system that treats them identically will book the wrong one first and lose the one that mattered.
- There is no way out to a human. The caller who most needs a person is the one already upset. Trapping them in a menu tree with no escape is how a missed call becomes a bad review.
- It pretends to be a person. Contractors mention this repeatedly and they are right to. The moment a customer works out they were fooled, the goodwill is gone and the deception is what they remember, not the convenience.
- It was sold as a replacement for staff. It is not one. It is a replacement for voicemail. Vendors who oversell that are the reason the whole category now carries a smell.
The comparison that actually applies to you
Here is where we think the trade forums get it slightly wrong, and it is worth saying even though it argues for our own product.
The hostile comparison is usually AI against a good receptionist who knows your customers and your schedule. On that comparison, AI loses, and it should. But most small contractors are not choosing between those two things. They are choosing between a machine and nothing at all, because at 9pm on a Sunday there is no receptionist, and the honest alternative is voicemail that the caller will not use.
One owner described the corner precisely. He could not justify forty thousand a year for an office manager, and the call centres he had tried were, in his words, useless. He asked whether there was any middle ground, and said he felt he was hitting a growth ceiling purely because he could not answer the phone.
That is the real question. Not whether AI beats a great human, but whether it beats the silence that is currently there.
Four things that have to be true
If you are going to put anything between your customers and your phone, these are the conditions we would insist on, whoever you buy from.
- It answers on the first or second ring, every time. The entire reason for its existence is that it never gets busy. If it is slow, the human alternative is better and you should buy that instead.
- It recognises an emergency and treats it differently. No heat, no water, active leak, sparking panel, storm damage. These should trigger a different path and, if you want them to, ring your phone directly regardless of the hour.
- A caller can reach a human by asking. Not through a menu. By saying so. And the handoff needs to be fast enough that they do not give up while it happens.
- It never claims to be a person. If asked, it answers honestly. In our experience the honesty costs almost nothing, because the caller wanted their appointment booked, not a friend.
When you should not buy this
If you answer your own phone and rarely miss it, this solves a problem you do not have. Buy nothing. Spend the money on the thing that is actually constraining you.
If your jobs are large, rare and relationship-driven, where a single call might be worth tens of thousands and the customer expects to speak to the owner, put a person on the phone. The economics justify it easily and no machine is going to build that rapport.
If you already have someone good answering during business hours and your after-hours volume is genuinely thin, the honest answer is that you might only need better voicemail handling and a follow-up habit, not a system.
Where it does earn its place is the specific and very common situation of a growing shop: the phone rings while everybody who could answer it is under a house or on a roof, calls arrive after five and at weekends, and the owner already knows he is losing work but cannot yet justify a full-time hire.
Measure it before and after
Whatever you decide, do not buy on a demo. Demos are performed with a cooperative caller and a quiet room, which is nothing like a Sunday in August.
Pull thirty days of call logs first and count what you are missing now, including how many of those missed calls you eventually recovered. Then run whatever you buy for another thirty days and count three things: how many calls it answered, how many turned into booked jobs, and how many callers hung up.
If the hang-ups outnumber the bookings, it is not working, and no amount of vendor reporting should convince you otherwise. That is the same test the owner in the seven-call experiment applied, and it is the right one.