AI Receptionist Integration for ezyVet: How It Works for Veterinary Practices

A dog swallows something at 7:40 on a Tuesday evening. The owner is standing in the kitchen with the phone in one hand, already scrolling to the next clinic in the search results. Your practice closed at six. The call hits voicemail, and you find out about it the next morning, if you find out at all.

ezyVet runs the practice well. Patient records, invoicing, inventory, the appointment book, the clinical history that follows a patient for its whole life. It does what a practice management system is supposed to do. It does not answer the phone.

That gap is where most veterinary practices lose money, and it is not only after hours. It is the Monday morning rush when every line is lit and the front desk is checking in a nervous cat. It is the lunch hour when nobody is at the desk. It is the four minutes a receptionist spends walking a caller through the boarding policy while two other owners give up and call the practice down the road. Every one of those is a client who was ready to book and did not get the chance.

Advisory Formulas built an AI receptionist integration for ezyVet to close that gap. This post covers what it does, what it deliberately will not do, and what it has captured in production.

What was built

The client is a small-animal veterinary practice in the Charlotte metro area that has been serving the area for almost 30 years. They run ezyVet as their practice management system and had the ordinary problem: a phone that rings more than two people can answer, and an appointment book that only staff could write to.

Advisory Formulas deployed a bilingual AI receptionist, English and Spanish, on the practice's main line. She answers every inbound call, identifies the caller and the pet, answers routine questions from the practice's own knowledge base, takes callback requests, transfers anything urgent to a person, and books routine appointments directly onto the practice calendar in ezyVet.

That last part is the piece most answering services cannot do. A booked appointment appears on the correct provider's column in ezyVet with the correct appointment type and a note marking it as booked by the AI receptionist. Nobody retypes it. Nobody copies it over from a message pad in the morning.

Call answering went live August 2026. Booking into ezyVet went live the week of September 21, 2026. The booking integration took more than 70 hours of build and test work across several weeks, most of it spent on correctness rather than wiring: making sure an existing client is matched to their existing record rather than duplicated, making sure the right pet is attached to the right appointment, and making sure a failure never results in a caller being told they are booked when they are not.

On the vendor relationship, the exact and literal status: the integration was submitted to IDEXX, the maker of ezyVet, through their certification process. IDEXX certified it and promoted it to their production environment. Advisory Formulas retains ongoing sandbox access for testing. IDEXX did not build, review, or endorse the AI receptionist itself. The certification covers the integration against their API.

A real call, start to finish

It is 6:50 on a Thursday evening. The practice closed at six. A woman calls because her terrier has been scratching at one ear for two days and now there is a smell.

The AI receptionist picks up on the first ring and greets her by the practice name. The caller gives her name and her dog's name. The receptionist recognizes her as an existing client from the phone number she is calling from, so it does not make her spell anything she has already given the practice before.

She asks whether this is an emergency. The caller says no, the dog is eating and acting normal, it is just the ear. That answer matters. If she had said the dog was in distress, or had simply asked for a person, the call would have transferred immediately. End-of-life and emergency calls never get handled by a script.

So the receptionist offers to get her on the book. The caller asks for Monday morning. The receptionist confirms the pet, the reason for the visit, and the time, repeats it back, and tells her she is set for Monday.

That appointment is now on the provider's column in ezyVet, under the right client record, attached to the right pet, flagged as booked by the AI receptionist. The front desk sees it when they open the schedule Friday morning. No voicemail to return. No callback to chase. No chance for the caller to keep scrolling to the next clinic.

Now the same call with a different ending. The caller asks about boarding her dog over Thanksgiving. Boarding is not something this practice wants booked automatically, and neither are grooming, surgery, sedation, or ultrasound. The receptionist does not improvise. She takes the request and the number, tells the caller the team will call her back, and logs it. The practice decided which services an AI is allowed to book, and the system holds that line without exception.

She also will not quote prices. Not an estimate, not a range. Price questions go to the team.

Results

Every number below is pulled directly from the call platform that runs the AI receptionist. Real calls, real periods, no projections.

Since going live on August 1, 2026, the AI receptionist has answered 298 calls for this practice. That is 67 days at an average of 4.5 calls a day. The average hides the part that matters.

Daily call volume rose 161% between August and early October. August averaged 2.8 calls a day. September doubled it to 5.6. The first six days of October are running at 7.3. The busiest single day so far was October 2, with 27 calls. The busiest week was the week of September 7, with 62.

Read that curve carefully, because it is the whole argument. The receptionist did not get a one time bump and settle. Month over month, more of the practice's phone traffic is landing on a system that picks up every time. The practice is not suddenly fielding more calls than it used to. It is capturing more of the ones it was always getting.

Average call length dropped 20% over the same period. August calls ran 1 minute 29 seconds. By early October they were down to 1 minute 11 seconds. That is the opposite of what usually happens when volume climbs. It happened because the receptionist's prompts and knowledge base were rewritten against transcripts of real calls, so callers stopped being asked for information the practice already had on file.

More calls handled, each one faster. That is the pattern a practice owner should look for when evaluating any answering system, and it is the pattern this one has produced for two straight months.

What that means at the front desk. At 7.3 calls a day averaging 1 minute 11 seconds, the receptionist is absorbing roughly 8 to 9 minutes of live phone time every day that a staff member would otherwise be holding, on top of every call that arrives when nobody is at the desk. Across a month that is hours of front desk attention returned to the animals in the building.

Nobody misses a call on the day the clinic is closed. Forty one of the 298 calls, 13.8%, arrived outside staffed hours. The single largest source is Wednesday, a day this practice is fully closed. Sixteen calls came in on Wednesdays, every one of which would have reached a recording and nothing else. Monday mornings before eight and evenings after six account for nine more. All of them were answered.

Twelve calls, 4% of the total, logged no conversation at all. Hang ups, silence, and robocallers that did not engage. All twelve landed during business hours, which is what robocaller traffic looks like. That is twelve interruptions a staff member did not have to absorb.

The receptionist handled 223 of 298 calls, 74.8%, without transferring to a person. The remaining 75, 25.2%, were escalated to the team. That split is the number to scrutinize in any AI receptionist, and the detail underneath it matters more than the headline: transferred calls averaged 54 seconds, against 88 seconds for calls handled end to end. The receptionist is not holding callers hostage while it fails. When a call is outside its scope, it hands off quickly.

Volume alone is not the test. Restraint is. In a detailed review of the first 195 of those calls, from August 1 to September 14, sixteen grooming and nail trim requests and seventeen boarding requests came in. Every one was routed to a callback rather than booked, exactly as the practice instructed. Twenty five calls arrived during the practice's midday closure, when nobody is at the front desk. A system that books what it was told not to book is worse than no system at all.

No revenue figure is published here. Attributing revenue to captured calls would require the practice's average visit value and a verified count of visits that actually happened, and that has not been measured. Call volume carries this post on its own.

What stays in ezyVet

Everything. ezyVet remains the system of record. The AI receptionist writes into it and reads from it. It does not replace it, mirror it, or sit in front of it.

There is no second dashboard for your team to check. There is no new calendar to keep in sync with the real one. Your front desk opens ezyVet in the morning the same way they did before, and the appointments the AI receptionist captured overnight are already there, on the right column, under the right client and patient.

That matters more in veterinary than in most industries. A patient record in ezyVet is a clinical record. Vaccination history, chronic conditions, medication, prior visits. Anything that creates a duplicate client, attaches a visit to the wrong animal, or splits one household across two records is not an inconvenience, it is a data integrity problem that a technician has to untangle later. Matching an existing caller to their existing record, and the right pet to the right appointment, was the single largest share of the build time.

The front desk also keeps final control of the book. Reschedules and cancellations go to a person, by design. The team can move anything the AI receptionist booked, and nothing is locked.

Who this fits

This fits a veterinary practice running ezyVet that is losing calls it can measure. You know the signs. Voicemails you return the next day. A front desk that cannot pick up during the morning rush. Spanish-speaking clients you lose because nobody on shift can take the call. Boarding and grooming questions eating the receptionist's afternoon.

It fits best where the practice has clear rules about what may be booked and what may not. The practices that get the most out of this are the ones with an opinion: this appointment type yes, these services always a callback, emergencies always a person. Clear rules make a reliable system.

It does not fit everyone, and here is who should pass.

If your call volume is low enough that your front desk picks up every time, you do not need this. Pay for it when the phone is actually costing you.

If you want an AI that handles complex scheduling, moves appointments around, or triages medical questions, that is not what this is. Reschedules, cancellations, and anything clinical go to your team.

If you are not on ezyVet, this specific integration does not apply, though Advisory Formulas builds the same pattern against other platforms. See the integrations page for the current list.

And if your practice does not have settled rules about what an AI may book, settle them first. A system that is told to use its judgment on clinical scheduling will eventually use it badly.

Hear it answer your phone

The fastest way to judge an AI receptionist is to call one. Advisory Formulas will set up a live demo on a line you can call yourself, using your practice's hours, services, and booking rules, before you commit to anything.

If you run ezyVet and the phone is costing you appointments, book a call and we will walk through what the integration would look like for your practice.

Book a 15 minute call or reach Advisory Formulas at 828-624-6891.

More platform integrations: see the full list.