We delivered an AI triage and client communication coordinator for a three-doctor veterinary clinic — fielding the after-hours flood of worried pet-owner messages, separating true emergencies from wait-until-morning cases, and running post-visit follow-ups the staff never had time for

A three-doctor veterinary clinic in West Ashley was drowning in communication, and it was costing them in every direction at once. Every morning, the front desk opened to forty-plus voicemails and portal messages that had accumulated overnight — a panicked owner whose dog ate a grape at nine p.m., someone whose cat had been vomiting for three days and finally decided midnight was the moment to act, a dozen refill requests, appointment changes, and at least one genuine emergency that had needed an emergency hospital eight hours ago and instead sat in a voicemail queue. The two front-desk staff spent their first ninety minutes triaging the backlog while the phones rang with the morning wave on top of it. Worse, the practice manager knew things were falling through the cracks on the back end: discharge instructions went home on a printed sheet that ended up on the floor of the car, post-surgical patients were supposed to get a check-in call at forty-eight hours that happened maybe half the time, and chronic-condition patients — the diabetic cats, the kidney-disease dogs, the seniors on long-term meds — quietly drifted away between visits because nobody had bandwidth to notice they were overdue. We built the clinic an AI triage and client communication coordinator, designed with the medical director around one hard boundary: the AI never diagnoses and never gives medical advice. It classifies, routes, gathers, and communicates — the clinical judgment stays with the doctors. The after-hours layer is where it earns its keep first. When an owner calls or messages at night, the AI responds immediately and works through a structured triage conversation the doctors wrote and approved — species, symptom, onset, severity signals, ingestion specifics. It knows the difference between one grape eaten by an eighty-pound Lab and a bag of raisins eaten by a ten-pound terrier, because the doctors encoded exactly which scenarios are red-flag categories. True emergencies get an unambiguous, immediate answer: this cannot wait, here is the emergency hospital, here is the address and phone number, go now — and the clinic gets a log so the doctor can follow up in the morning. Everything that can safely wait gets a calm, structured response — your concern is noted, here is what to watch for overnight per our doctors' standing guidance, you are booked into tomorrow's first urgent slot — and the owner goes to bed reassured instead of spiraling on internet searches at two a.m. By morning, the front desk no longer opens to an undifferentiated pile of forty voicemails. They open to a sorted queue: three urgents already booked into the day's held slots with symptom summaries attached, refill requests batched for technician review, appointment changes already handled, and everything documented in the practice management system against the right patient record. The ninety-minute morning dig-out became a ten-minute review. The follow-up layer closed the cracks on the back end. Every discharge now triggers a sequence the doctors configured by case type: surgical patients get their forty-eight-hour check-in automatically — how is the incision, is she eating, any lethargy — with the owner's answers summarized for the technician, and concerning answers flagged for a same-day doctor callback. Dental patients get their own sequence. New-puppy visits get vaccine-series reminders timed to the actual schedule. Discharge instructions go to the owner's phone as clear, readable text — no more crumpled sheet in the car — and owners can ask the AI clarifying questions about the written instructions they were already given: how much of this medication, with food or without, when can he go back to daycare. Anything beyond the written instructions gets routed to staff rather than answered. The chronic-care layer watches the population nobody had time to watch. The diabetic cat whose recheck lapsed, the senior dog eleven months past his last bloodwork, the kidney patient whose prescription food orders stopped — the AI surfaces them weekly with a suggested outreach message the staff approves and sends in one click. In the first ninety days, the clinic booked over sixty lapsed chronic-care patients back in — revenue that was walking away silently, and medicine that needed to happen. The after-hours triage handled more than four hundred conversations in the same period, sent nineteen true emergencies to the ER hours faster than the voicemail queue would have, and the post-surgical check-in completion rate went from roughly half to ninety-eight percent. The practice manager's summary: the phones are quieter, the mornings are sane, and for the first time the clinic follows up like the kind of practice they always believed they were.