An AI agent for veterinary practices: the admin, never the clinical bit

A veterinary practice runs on two entirely different kinds of work happening in the same building. There’s the clinical work, which is why everyone trained for years. And there’s the surrounding tide of reminders, forms, claim paperwork and phone calls, which is why reception never sits down and why the vet is doing admin at seven in the evening.

Only one of those is safe to hand to a machine. Fortunately, it’s the one eating the day.

What an agent can genuinely take on

• Appointment confirmations, reminders and no-show chasing. The quiet profit leak in any appointment book. A reminder the day before, a nudge to rebook when someone doesn’t turn up — relentless, polite, and never forgotten on a busy Tuesday.
• Vaccination and health-check recalls. Boosters, annual checks, parasite treatments. This is the classic case: hundreds of small, dated, entirely predictable prompts that genuinely matter to the animal’s health and that nobody has time to work through by hand.
• Insurance claim paperwork. Chasing the missing form, the history the insurer wants, the client signature nobody remembered. None of it is clinical; all of it delays a client getting their money back and your practice getting paid.
• Repeat-prescription requests. Taking the request, gathering the details, putting it in front of a vet to authorise. The agent handles the queue; the veterinary surgeon makes every decision about medication.
• The routine end of the phone and inbox. Opening hours, prices, parking, how to register, whether you’re taking new clients, where the out-of-hours service is. It’s a large share of calls and none of it needs a nurse.
• Registration and record-keeping admin. New client details, address and microchip changes, the tidying-up that otherwise happens never.

The line it must never cross

Everything above is administrative. What follows is not, and this is where a badly built system does real harm:

• Never triage. Whether an animal needs to be seen right now, this evening, or next week is a clinical judgement. An agent must not form a view, and must not sound like it has one. Anything hinting at a poorly animal goes to a person immediately — not after a clarifying question.
• Never advise, diagnose or dose. No “that sounds like it might be…”, no reassurance, no guidance on medication or amounts. Ever.
• Never handle the hard conversations. End of life, a difficult prognosis, a complaint, a bill someone can’t pay. A person, every time.
• Never let the emergency route be slower. If out-of-hours cover is involved, the fastest path to a human must be the most obvious thing in the whole system.

That is the same discipline we apply everywhere: an agent that knows precisely when to stop and fetch a person is the only kind worth having near a practice.

Will clients mind?

For the routine things, no — provided you say plainly that it’s an AI and a person is one message away. Most people would rather confirm an appointment or request a repeat prescription without waiting on hold. What they will never forgive is being held at arm’s length while they’re frightened about their animal. Get that handover instant and obvious and the rest is genuinely welcome; and being upfront that you’re using AI costs you nothing and buys a good deal of trust.

The honest summary

An agent won’t make your practice better at medicine. It will mean the recalls actually go out, the claim forms stop drifting, and the phone rings less for things nobody needed to ring about — so the people who trained to treat animals spend more of the day doing it. That’s how we run our own operation: the agent does the legwork, and the moment something matters, a person picks it up.

General information, not veterinary or regulatory advice — your professional obligations remain entirely yours.