A salon phone is not a generic small business phone, and that is why the generic demos look better than the real thing.
Half of what a groomer gets asked cannot be answered without seeing the dog.
Where we stand. We build pet business software and we are building an AI assistant of our own, Grazi. It is a published closed beta and you cannot switch it on, so nothing here is me selling you ours. Some of the honesty below is expensive for us to write down.
What it can genuinely do
These are the jobs where the technology is not the limiting factor.
Answer the question you have answered four hundred times. What do you charge for a cavoodle, do you do cats, where do I park, how long does a full groom take, do I need to be there. Fixed answers, written once.
Read live availability out loud. Do you have a Saturday. This is the single most common call in a grooming salon and it is pure diary lookup.
Take a booking, move one, cancel one. Inside rules you set, with the pet and the service attached, at 2am.
Work at volume without queueing. Six callers at once on a Monday morning after a long weekend is the scenario where a human answering service starts putting people on hold and an AI does not notice.
Chase the thing you never get round to. The eight-week rebook nudge. The client who has not been in since March.
Stay polite at 5:50pm on a Friday. Not a small thing.
What it cannot do
This list is longer and it matters more, because every item on it is a way the same product loses you money rather than saving it.
It cannot look at the coat. A matted double coat and a well-kept one are the same words on the phone and a two-hour difference on the table. If the AI quotes the well-kept price, you either eat the difference or you have an argument at pickup that you started.
It cannot price what it has not seen. De-matting surcharges, behaviour handling, a second groomer for a big dog. These are judgements made with a dog in front of you. An AI asked to quote them will produce a confident number, because producing confident numbers is what it does.
It cannot decide whether a dog is safe on your floor. Reactive, entire, unvaccinated, recovering from surgery, elderly. Someone who knows your setup has to make that call, and an AI that books a reactive dog into a shared bay at the same time as a nervous puppy has made a welfare decision with no basis for making it.
It cannot read distress. A caller whose dog has just been stung, or who is crying because the vet has given them a prognosis, needs you. Ask any vendor exactly what triggers a handover to a human, and do not accept "it understands context".
It cannot notice what it does not know. This is the one that costs real money. An AI that has not been told you stopped taking poodles in June will keep taking poodles in June. The failure is silent. You find out at the door.
It cannot carry responsibility. If it promises a slot you do not have, you are the one ringing the customer back. The customer's contract is with you, not with the vendor whose software said it. Configure accordingly.
The question that actually decides it
Not "is the AI any good". This one:
Could you hand your prices and policies to a new employee on their first shift and have them get it right without asking you?
If yes, an AI will probably work, because the same written rules are what you configure it with.
If your pricing lives in your head, if the Saturday surcharge is "depends who it is", if the breed list is "I know the ones I do not take", then an AI cannot learn it either. It will invent something plausible instead. You do not have an AI problem at that point, you have an undocumented business, and writing it down is worth doing whether or not you ever buy anything.
Most salons I have met are in the second group. That is not a criticism. It is just the actual prerequisite, and no vendor will tell you it exists.
What to do about the gap
Split the phone rather than replacing it.
Let software handle the bookable and the repeatable, which is most of the volume and none of the judgement. Keep the judgement calls coming to a person, and give them a smaller, quieter phone to answer.
That split does not need an AI at all. A booking page with live availability handles "do you have a Saturday" better than an AI reading the same diary, because the customer sees every slot at once instead of being read three. Over half of appointments on Petboost are booked by the customer, and around 70% of those land outside opening hours.
The times are worth knowing before you price an alternative to them. A returning customer is through the booking form in about 38 seconds and a new one in 56, including their own details and the dog's. Nothing was metered and nobody was interrupted.
Where an AI earns its keep is the layer above that: the caller who will not use a booking page, the question that is not on your site, the message at 11pm that needs a reply rather than a form. That is a real job and it is worth paying for once the cheap layer underneath is doing its part.
If an AI answering your messages is the specific thing you are buying right now, there are products shipping one today and ours is not among them. Our comparison pages put the ones we are measured against next to us, prices included.
An AI receptionist for pet boarding and daycare
Boarding and daycare phones ring for different reasons, so both lists above shift.
"Have you got room from the 20th to the 3rd?" is a diary lookup across a date range. An AI can read the answer out, but a booking page answers it better: a stay books online inside your check-in and check-out windows, and the owner sees every free night at once. Drop-off times and the vaccinations you ask for have written answers either way.
What does not move to software is judgement. Whether a new dog can join the group is decided at an assessment, never over the phone. An entire male or a dog on medication is a decision about your floor and your staff, and a medication plan should be taken down by the person who will give it.
For daycare, the assessment gate does the vetting a receptionist would otherwise try to do on a call. Make the temperament test the only thing a new dog can book, and once a dog has passed, its owner books its days online.
Then there is the call that fills a boarding phone over the holidays: an owner who is away, asking how their dog is going. An AI on the phone cannot see the dog. A photo and a line from whoever did breakfast this morning can answer it, and in Petboost that is a customer-facing note with photos on the stay. When you publish it, Petboost can email the owner, and it opens in their app as a report card.
Where we stand here too. Our own AI for answering clients, the Grazi AI Booster, is in closed beta and you cannot switch it on. The dedicated number that comes with the Communications Booster is for texting: owners text it and you reply from the Messages inbox. With the Call Catcher Booster, an add-on you switch on yourself, it takes phone calls too. They ring your own phone, and a call you miss gets your booking link by text or goes to voicemail in Messages. One recovered call, or one booking link sent, a month pays for the whole $19. No AI answers it.
Before you sign
Ask for a recording of it handling a matted-coat quote, a reactive dog, and a caller who is upset. Not a demo script. A real one.
Running boarding or daycare? Add a holiday enquiry for dates you are already full on, and an owner who wants to skip the assessment.
If the vendor will not, that tells you where their product stops, which is the thing you were trying to find out.
What these cost in Australia · The three products compared · See self-service booking