Get the phone line right first: what a veterinary clinic should fix before it adds AI

For a while now the veterinary industry has been arguing about AI receptionists. Should a robot answer the phone? Will clients accept it? How far is too far? It is a fair argument. But the whole conversation skipped a step. Almost nobody asked whether the phone line underneath was any good.

That is the question we want to put first, because it decides more than the AI does. Most clinics run their front desk on a carrier line or a small-business VoIP plan that was never designed for a desk taking about 160 calls a day, as the practice in our own data did. Those systems cannot tell you how many calls went unanswered yesterday, how long callers waited, or how many gave up in the queue. Putting an AI on top of that does not fix it. It hides it a little longer.

So here is the argument in one sentence: the first thing a veterinary clinic should get right is a really good phone line, and AI should be a layer that makes that phone line better. This belief is why we built PupPilot as a veterinary phone system first — a full phone company with numbers, desk phones, and routing — rather than only as software that sits on top of whatever a clinic already has. (A clinic can still start our AI on its existing phones and port later. But the phone line is the product.)

What a premium phone line means for a veterinary clinic

"Telecom" is not a word most practice owners want to think about, so let's be concrete. A veterinary clinic is a medical office with a waiting room, and it should hold its phones to the standard of any medical office phone system. That standard has 4 parts, plus routing, which gets the rest of this piece.

Real numbers. Your main number is on the door, on every reminder card, and in the Google listing your clients tap. It should belong to you, port in cleanly when you switch, and be portable out again; the rules are in the checklist at the end and on our switching and porting page, because porting is where clinics get hurt.

Reliable service. A clinic phone system should keep working when the internet hiccups, fail over to cell phones when it does not, and tell you when something breaks. Ask any vendor what a caller hears when your building loses power; "a busy signal" is the wrong answer. Ours is on the reliability page.

Professional desk phones. A front desk is a physical place. A receptionist checking in a nervous dog needs a desk phone with a headset, a hold button, and a light that shows which lines are busy — not a softphone on a personal cell phone that logs out at the worst moment. The phone should belong to the desk, not to whoever left last month.

Unlimited calling and texting from the main number. Clients text the number they call. If your texting lives in a separate app with a different number, your team watches 2 inboxes and your phone system cannot offer "text us instead" to someone waiting in the queue. Texting from the main clinic number, with no per-message cap, is part of the phone line, not a separate product. That is how we built business texting.

How call routing should work at a veterinary clinic

Inbound call routing is the part of a phone system your clients actually experience, and most of it is configuration rather than technology. Any decent call routing software can do what follows; the difference is whether anyone set it up.

Start with the auto attendant. An auto attendant phone system exists to get a caller to a person, or to the thing they called for, in 2 prompts or fewer, counting greeting plus menu as 1 and a sub-menu as 2. The emergency instruction comes first, before the menu and before your hours, because the caller with a dog that cannot breathe is not going to wait through your Saturday schedule. Pressing 0 should reach a person from anywhere. We have published phone tree examples for veterinary clinics, with scripts.

Behind the attendant sit ring groups: the phones that ring together when a caller picks an option. Appointments should ring every front-desk handset at once, then roll to a second group, such as the treatment area, after a set number of seconds. Emergencies ring everything, ahead of the queue, and never end at voicemail during business hours.

When more calls arrive than the group can answer, they go to the call queue, and this is where most clinic phone systems fail. A queue should offer a way out that is not hanging up: a callback that holds the caller's place in line, or the option to text the same number instead. A caller who switches to text has not abandoned. The queue also needs a timeout, and the timeout should send the caller somewhere that answers, not into a mailbox.

Hold is the other place callers give up. A receptionist who asks before putting someone on hold, and checks back inside a minute rather than letting the hold run, keeps most of those callers. For scale on how long people will actually wait: in the 2025 JAVMA secret-shopper study, 8.2% of callers trying to book an appointment gave up after more than five minutes on hold.

Voicemail-to-text turns the messages that do arrive into something the desk can read between check-ins. It does nothing for the callers who hung up instead of leaving one, so voicemail should be the last resort in the tree, not the default after hours.

Finally, business-hours rules. The system should know when you are open, at lunch, or closed for a holiday, and route differently in each case without someone remembering to flip a switch at 6 p.m. After hours, the emergency number comes first.

Outbound routing: outreach that comes back to the right place

The other half of call routing gets less attention, and it is where a clinic quietly creates a second inbound mess for itself. Reminders, callbacks from the queue, refill-ready texts, post-surgery check-ins, lab results: a busy practice sends a lot of calls and texts.

Every one of them should satisfy 2 rules. First, it goes out from the clinic's own number, the one clients recognize, not from a short code or a number assigned by a separate reminder app. A client who gets a text from an unknown number ignores it, or calls the main line to ask who texted them: a new inbound call you created yourself. Second, the reply lands in the right place: a response to a refill-ready text with the pharmacy, a reply to a post-op check-in with the nurse who sent it, and "can I move that to Thursday?" at the front desk. When outreach goes out from one number and replies come back to a queue that knows what they were about, the desk works them between clients. When it goes out from 5 different tools, every reply is a mystery call.

What a medical office phone system should report

Here is the test we'd apply to any medical office phone system, veterinary or otherwise: can it tell you, for last week, how many calls were offered, how many a person answered, how many the caller abandoned, how long answered calls waited on average (the average speed of answer, or ASA), how long callers spent on hold, what share were answered within a set number of seconds (the service level), how many timed out of the queue, and how many arrived after hours? Call centers have reported those numbers for decades. Most clinic phone lines report none of them.

Without them you cannot know whether you have a problem, let alone whether AI would fix it. There is no independently verified veterinary abandonment benchmark, but there are honest reference points.

The closest published measurement of veterinary phone-layer failure is a 2025 secret-shopper study of 5,053 US practices in JAVMA: 15.1% of booking attempts could not reach a staff member and 8.2% ended after more than five minutes on hold. It measures new-client wellness calls during business hours rather than total call traffic, so it is not an abandonment rate — but it is peer-reviewed and it is the number we would use.

For contact-center guidance, SQM Group, a benchmarking firm, points in the direction of lower is better, and past roughly one call in ten you have a problem; published thresholds vary between sources, so we quote the direction rather than a single number. The closest thing to healthcare data is a peer-reviewed study of the US Department of Veterans Affairs call centers (Griffith et al., American Journal of Managed Care, 2019): across 285 facilities, average speed of answer improved from 87 to 69 seconds and abandonment fell from 12.0% to 8.3%. That is a very large health system working at it for 2 years, and it got to 8.3%, not 5%. The definitions and formulas are in our companion piece on call abandonment rate for veterinary clinics, and every benchmark above is documented — source, sample, and scope — on our verified statistics page.

One clinic, one week

The best evidence we have for all of this is our own platform data from one general practice, 5 weekdays before it moved its lines onto PupPilot and 5 weekdays after. It is one clinic and one week: a case, not a study. The full write-up is in the 35% problem.

Before the switch, the practice received about 160 inbound calls a day and its team answered about 105 of them. Over 5 days, roughly 280 calls — 35% of everything that rang — were never picked up by a person. And 35% is a floor, because the old system could only log the calls that reached it. Callers who met a full queue, timed out, or hung up before entering the queue never registered at all. The phones that would have measured the problem were the phones missing the calls.

Over an equivalent 5 days after the switch, the AI layer handled nearly 2,000 interactions on the same lines. More than 1,000 were transferred to reception, a nurse, or the pharmacy with the caller's name, patient, and reason for calling already on screen. About 900 were resolved on the line: questions, triage, and other routine needs, handled at the moments no one was free to pick up. Across all of them, 4.3% of callers asked not to deal with the AI and were moved to a person. The team did not handle fewer conversations. It handled more of them, and better ones, because the phone line was finally answering everything and counting everything.

Where AI fits, once the line is right

Now, and only now, the AI. PupPilot is a veterinary phone system — unlimited calling and texting on professional desk phones — with an AI front office on every line. The AI answers the calls your team can't get to — after hours, overflow, and when the desk is busy — handles the routine ones, and transfers the rest to your staff with the caller's name, patient, and reason for calling already on screen. It isn't a replacement for your receptionists. It's the part of the phone system that makes sure every call reaches someone.

In routing terms, the AI is the best possible timeout. Where the queue used to overflow to voicemail, it overflows to something that answers. Where the after-hours greeting used to end at a mailbox, after-hours coverage picks up, sends the urgent case straight through, and books the rest for the morning. Whatever needs a person comes back through a live transfer with context: the record from your practice information management system (PIMS) is on screen before your receptionist says hello, and the caller does not repeat themselves.

The handoff is the part we care about most. A caller who asks for a person gets one. A caller who sounds frustrated gets one without asking. Both are added to an opt-out list so their next call skips the AI. That is where the clinic's 4.3% came from. Platform-wide, across PupPilot clinics, the opt-out rate runs closer to 1.5–2% — the verified statistics page carries both figures.

This is the point of the whole piece: bolting an AI onto a bad phone line does not fix the phone line. If the queue has no callback, the AI inherits a queue with no callback. If texting lives in a separate app, the AI cannot text from your number. If the system cannot report answered rate, the AI's results are invisible too. The tools that sit on top of whatever phone system you already have can be good at answering a call; they cannot make the line underneath them better. That is why we became a phone company.

What to ask any vendor

If you are evaluating a phone system, an AI add-on, or both, 5 questions sort the field quickly.

Does it supply the phone lines, or sit on top of them? If you still need a carrier after you buy it, you are buying a layer, and the line underneath is still your problem.

Can you port your numbers in, and out again? Get the porting terms in writing. The FCC requires simple ports to complete within 1 business day and does not allow a carrier to refuse a port over an unpaid balance. Some vendors' terms only allow a port while the account is active, so port first, cancel second.

Is texting from the main clinic number included, without a per-message cap? Texting from a different number, or through a separate app, is not the same thing.

What does it report? Ask to see last month's offered, answered, abandoned, ASA, hold time, service level, and after-hours volume from a real customer. A vendor that cannot show those numbers cannot show you what it fixed.

What happens to a caller who wants a person? Ask about business hours, after hours, and the moments the AI is not sure. The answer should be a transfer with context — the PIMS record already open, not a mailbox — and the caller should not have to fight for it.

Our own answers, alongside the other systems clinics consider, are in the guide to the best veterinary phone systems. Whatever you choose, get the phone line right first. Everything else, including the AI, is only as good as the line it runs on.