Every veterinary practice manager knows the feeling of walking past the front desk at 9:40 on a Monday morning: three lines lit, one receptionist checking in a nervous dog, another on hold with a specialty hospital, and the phone ringing anyway. What almost nobody knows is the number behind that feeling — how many of the calls that came in today were never answered by a person at all.
We recently looked at 5 weekdays of phone data from a multi-doctor general practice before and after it moved its phone lines onto PupPilot's veterinary phone system. The numbers are from one clinic and one week, so treat them as a case, not a study. But they are the clearest picture we have seen of what a busy clinic's phone line is actually doing, and they changed how we talk about our own product.
Before: about 160 calls a day, about 105 answered
In the 5 days before the switch, the clinic averaged about 160 inbound calls a day. Its team answered about 105 of them. Over the week that came to roughly 280 calls — 35% of everything that rang — that nobody on the team picked up. Those callers got voicemail, gave up on hold, or hung up.
The 35% is an undercount, and this is the part worth sitting with. A phone system can only log the calls that reach it. When the queue is full, when a caller listens to the hold music for 90 seconds and drops, when someone dials, hears the auto-attendant and decides to try again later, those attempts either never register or register as nothing more than an abandoned call with no name attached. The clinic's real unmet demand was larger than 35%; the phone system simply had no way to show it.
That is the first lesson, and it has nothing to do with artificial intelligence: a clinic cannot manage what its phone system cannot measure. Most practices are running on a carrier line or a small-business VoIP plan that reports minutes used and not much else. Answered rate, abandonment rate, average hold time, queue timeouts — the four numbers that describe whether clients can reach you — are invisible.
After: five days with AI on the line
Over an equivalent 5-day period after the switch, PupPilot handled nearly 2,000 interactions on the clinic's lines. More than 1,000 were transferred to staff — to reception, a nurse, or the pharmacy — by live call transfer, with the caller's name, the patient, and the reason for the call already on screen. About 900 were resolved on the line — the frequently asked questions, emergency triage that routed the urgent cases straight through, and the other call needs that never needed to reach reception.
Two things stand out in that split.
The team did not handle fewer conversations. It handled more, and better ones. Before the switch, staff picked up just over 500 calls in a week. After it, more than 1,000 calls reached a person — but they reached that person after the AI had already asked why they were calling and pulled up the chart, and without the caller having to repeat themselves. The receptionist's morning is no longer a queue of unknown calls; it is a set of conversations that needed her.
The volume itself is the other tell. Nearly 400 interactions a day against about 160 calls a day before is not the AI inventing demand. It is the demand that the old phone system was hiding — the callers who used to hit a full queue and hang up, the after-hours calls that used to go to a voicemail box nobody could act on until morning, and the clients who now text the clinic's main number because they finally can.
The question everyone asks: what about the clients who hate it?
Some did. A few were rude about it. PupPilot is built so that a caller who shows frustration, or simply says they do not want to talk to an AI, is moved to a human and added to an opt-out list automatically; from then on, their calls skip the AI layer. Over the same 5 days, 4.3% of interactions ended in an opt-out. The other 95.7% stayed on the line and got what they called for.
We think 4.3% is the honest number to publish, not a smaller one. Change on a clinic phone line is felt by every client, and the right design assumption is that some people will always want a person. The job of the AI is not to argue with them. It is to get them to that person faster than the old hold queue did, and to keep the line clear for everyone else while it does.
What this means for how a clinic should think about its phones
It is tempting to read a case like this as an argument for automation. We read it differently. Almost everything the clinic gained came from things a good phone system should have been doing already: answering every call, measuring what happens to it, routing it to the right person, and giving that person context before they say hello. The AI layer is what made those things possible at 9:40 on a Monday and at 11:15 on a Saturday night, when no one is free to pick up. It extends the team's coverage; it does not replace the team.
If you run a practice, the place to start is not a demo of anything. It is 3 numbers you probably cannot get from your current phones:
- Answered rate — of the calls that rang today, how many did a person pick up?
- Abandonment rate — how many callers hung up while on hold or in the queue?
- After-hours volume — how many calls arrived when nobody was there, and what happened to them?
Once those numbers exist, the conversation about AI becomes a much simpler one: here is the 35% (or 20%, or 50%) that nobody answered last week, and here is what it would take to answer it. How to compute them from a carrier export or a tally sheet, and what a good number looks like, is in our piece on call abandonment rate for veterinary clinics. And when you want to compare against what has actually been measured — rather than the numbers that merely circulate — the verified statistics page holds every benchmark we trust, with its sample, method, and limits.