This is the last of a three-part series on AI in the service industry. Part 1 covered the research. Part 2 covered what pet businesses actually ask AI to do.
Where we stand. We build Petboost, booking and business software for pet businesses, and the second half of this post is about our own product. The first half is a ladder you can use whatever software you run.
At the start of October the vet profession got something pet businesses do not have yet: a proper measurement of how deeply AI sits in its clinics. A veterinary software company and the American Animal Hospital Association surveyed 1,730 vet professionals across the US and Canada. The headline was adoption, which more than doubled in two years, from 39.2% to 83.7%.
The finding I keep coming back to is a different one. Nearly two in three practices (65.2%) added AI without changing how they work. Nearly half have no plan for it and nobody responsible for it. And the practices the study rated most mature reported roughly six times the outcomes of the least mature. About one in five users reported no measurable impact at all.
In the study, the stronger outcomes went with the practices that had changed how the work runs, not only added a tool to it.
Why the vet numbers matter to a groomer or a daycare
Vets are a useful mirror because they sit next to you. The clients are the same people, the animals are the same animals, and the front desk does the same juggling.
The top use in clinics was writing notes, through AI scribes that listen to a consult and draft the record. Client messages came second. Reliability worried 67.4% of users and data security 59%, and both are the right worries. Fear of losing a job had fallen to one in five overall, but it was still 37.8% among receptionists, which tells you where people feel AI is aimed.
None of that is a reason to rush. It is a reason to be deliberate. Adding a tool beside the old way of working was the most common pattern in the study, and the stronger results sat with the practices that went further.
The ladder
We rewrote the maturity idea for the jobs a pet business actually does. Every job sits on one of five rungs.
0. By hand. You do it yourself, or it does not get done.
1. Copy and paste. Someone pastes the job into a chatbot and copies the answer back. Useful, but it lives in one person's head.
2. A separate AI tool. A tool does the job well, outside your booking system, so you are still the bridge between it and your records.
3. Built in. AI inside the system where your bookings and clients live does the job, and a person checks it before it reaches anyone.
4. Runs within rules. AI does a narrow job on its own, inside rules you wrote, logs everything, and hands over to a person when it is unsure.
The step that changes the most is from 2 to 3. Below it, AI only knows what you pasted in. Above it, it can see the booking, the pet, the history and the takings, and the answers stop being guesses.
Level 4 is not a target for everything. A booking that involves money, a health condition or a dog that needs two handlers can sit with a person forever, and that is a sound decision rather than a slow one.
Six guardrails to put in first
Before any job moves past level 2, get these in place. They are short, and most of them are a conversation rather than a purchase.
- Written rules for what staff may put into AI tools, and who decides on new ones. Our AI rules for your team post has a template to start from.
- A person reads anything AI writes before a client sees it.
- Clients are told when they are dealing with AI, the first time and not only when they ask.
- Bookings, clients and pets live in one system. AI built on three spreadsheets gives confident answers from half the facts.
- One named person owns how the business uses AI.
- Clients can always reach a person, and someone actually picks up.
The first three line up with what Australian small businesses told the National AI Centre: about half check AI output before it reaches a customer, and telling customers lags well behind. The last one lines up with what customers say they want almost everywhere it has been asked.
Score yourself
We turned the ladder and the guardrails into a free AI readiness assessment for pet businesses. Eight everyday jobs, six guardrails, a level from 0 to 4 and your next three moves. It asks for no email, and nothing you pick leaves your browser.
One rule in it is deliberate. If fewer than four guardrails are in place, the result is held at level 2 however advanced your tools are. Handing work to AI without them is not maturity.
There is no comparison with other businesses in the result, because nobody has surveyed pet businesses properly yet, and we would rather leave the line blank than invent an average.
How this maps to Petboost today
Most of what we have built so far sits at level 3 on purpose.
Grazi, the assistant included in every Petboost plan, reads your live bookings, clients, pets and takings, answers questions about them, and drafts descriptions, notes and replies in your voice. It cannot book, cancel, charge or send. You read what it drafts, and you press send.
That is the right rung for most of the jobs in Part 2: the Monday takings question, the list of dogs who have not been in for eight weeks, the service description written at 9pm.
The next three things we are building move a few narrow jobs further up the ladder, and each one keeps the guardrails above.
Coming next: talk, and it is written down
The best notes in a pet business are the ones nobody has time to type. The groomer knows the dog flinched at the back legs. The walker knows which gate sticks. By the time anyone sits at a keyboard, half of it is gone.
We are working on voice notes inside Grazi. A team member talks for twenty seconds at the end of a visit, and Grazi writes it into that pet's timeline as a tidy, dated note. Recording and transcribing calls on your business number, for training and for a record of what was agreed, is on the public roadmap as well.
Two rules carry over from how Petboost handles notes today. Transcribed notes land internal by default, visible to your team and not to the client. And you can always see what was said and what was written, side by side.
Coming next: themes
Report cards are the one thing in a pet business that clients reliably open, screenshot and share. We are designing report card themes: seasonal looks, occasions such as a birthday or a first visit, and milestones such as a dog's tenth groom, picked from the date, the service and the pet's own history so nobody has to remember to choose one.
The other kind of theme is the one owners keep asking Grazi for by hand: what came up most this week. Which questions clients kept ringing about, which services they asked for that you do not list, what the team kept writing in notes. Summarising those themes into one short read is the natural next step from the Monday questions people already ask.
Coming next: when nobody can pick up
This is the one owners ask about most, and the one we are taking slowest.
When your hands are wet, or you are mid-clip, or the van is on the highway, the phone still rings. The Grazi AI Booster is your own assistant, answering on your business number when your team cannot. It is a two-way conversation, not a voicemail: it greets the caller in your business's name, with your greeting and a voice you pick from a shortlist, answers from your own records, and can book, move or cancel inside the booking rules you already set.
It shows up in Petboost as a team member, so everything it does is logged against it. Every action sits behind its own switch, and the ones that move money start switched off. The most it will do with money is send a payment link. It never charges a card on file on its own. It says it is an assistant when asked, and it hands over before it guesses: money, health and complaints go to a person, with the conversation waiting in Messages.
It is in closed beta and still being built, so you cannot switch it on today, and there is no price or date yet. If you want to be told when it opens, the beta list is on the Grazi AI Booster page.
We are going slowly for the reason the research in Part 1 keeps pointing at. Customers will give AI a go when it helps them and they can reach a person. For our own assistant, being right most of the time is not good enough yet on a line where a wrong answer means a dog turning up on the wrong day. When it is, it will sit at level 4 for a few narrow jobs, behind every guardrail on this page.
Sources
- American Animal Hospital Association with Association Insights and a veterinary software company, The State of AI in Veterinary Medicine: 2026, published October 2026. 1,730 vet professionals in the US and Canada, surveyed July to August 2026.
- National AI Centre, AI adoption insights: December 2025 to February 2026. ai.gov.au.
- Gartner, survey of 3,566 customers on GenAI in customer service, August 2026. Press release.
- Petboost's own product pages for Grazi and the Grazi AI Booster, read at the time of writing.