Safe by design: how Tilly handles health calls
Pet owners call their vet when they're worried. Tilly is built around one idea: an AI should never stand between a worried owner and a person who can help. It doesn't give advice, it doesn't judge how serious something is, and when it hears signs of a possible emergency, it gets the caller to a person straight away.
Updated
No medical advice.
Tilly never diagnoses, names medicines or doses, or suggests home remedies.
Emergencies go to a person.
A possible emergency is passed to your team, or your emergency partner after hours, within seconds.
Always says it's an AI.
Every call and message starts with "I'm Tilly, the AI assistant at your clinic."
Three rules that every call follows
Tilly never gives medical advice.
No diagnoses, no medicine names, no doses, no home remedies, no reassurance that something isn't serious.
Possible emergencies go to a person, at once.
Every sentence a caller says is checked for red flags before Tilly does anything else. If in doubt, Tilly treats the call as urgent.
Callers always know it's an AI, and can always reach a person.
Every conversation starts with Tilly saying it's an AI. Anyone who asks for a person is offered a transfer or a callback, every time.
These rules are enforced in software, outside the AI model. The model cannot switch them off, and a caller can't talk Tilly out of them.
What Tilly says, and what it won't
Veterinary advice belongs to veterinarians. Tilly checks every sentence it's about to say, on every channel, before it's spoken or sent. If a sentence would cross the line, it's replaced with: "I'm not able to give medical advice, but I can get you to someone who can."
Tilly will
- Recognise signs of a possible emergency and connect the caller to people
- Give your clinic's hours, address and logistics from answers you approved
- Give your emergency partner's name, number and address
- Book an appointment so a vet can see the pet
- Take a message or request for your team
- Pass photos to your team without looking at them
Tilly won't
- Diagnose or guess what's wrong ("sounds like bloat")
- Name medicines or substances as treatment
- Give doses
- Suggest home remedies or first aid (such as making a pet vomit)
- Say something is "probably fine" or "not serious"
- Predict how things will turn out
- Interpret test results, symptoms or photos
Example from our demo clinic
- Caller
- Can I give my dog ibuprofen for his leg?
- Tilly
- I'm not able to give medical advice, but I can get you to someone who can. Would you like me to book Max in, or ask a vet to call you back?
What Tilly listens for
Tilly works from a list of red flags written and approved by licensed veterinarians. It listens for them in what callers say and what they write, in every supported language, and it takes the type of animal into account.
- Trouble breathing, choking, blue or grey gums
- Collapse, fainting, can't stand, won't wake up
- Seizures or shaking that won't stop
- Heavy bleeding or bleeding that won't stop
- Suspected poisoning: chocolate, sugar-free gum, rat or slug bait, human medicine, antifreeze, grapes or raisins, cannabis edibles, and lilies for cats
- A swollen, hard belly with retching or trying to vomit without bringing anything up
- Straining to pee or not peeing, especially male cats
- Hit by a car, a fall from a height, an attack by another animal
- Problems giving birth
- Eye injuries or sudden blindness
- Overheating or being left in a hot car, with extra care for flat-faced breeds
- Vomiting blood or blood in stool
- Severe pain: crying out or not letting anyone touch them
- Rabbits or guinea pigs not eating or not passing droppings for 12 hours or more
- Birds sitting on the cage floor or bobbing their tail to breathe
- Puppies and kittens under 12 weeks who are vomiting, have diarrhoea, won't eat or seem flat
- Anything the caller calls an emergency
When in doubt, it's urgent. If the caller says something that might be a red flag, Tilly treats it as one. If they describe a symptom and ask whether it's serious, that's urgent too. Once a conversation is urgent, nothing later in it, including the caller saying "actually, he's fine", can make it less urgent.
This list is maintained by Tilly's clinical advisors and may change. Your clinic can add red flags of its own.
Concerned but not urgent: If a caller is worried but there's no red flag, Tilly carries on helping, offers the earliest appointments first, marks any task high priority, and tells them: "If anything changes or you're worried it's getting worse, please treat it as an emergency and call us, or Riverside Emergency Vets at 555 014 2200 if we're closed." (demo clinic example; your emergency partner's details are used)
What happens when Tilly hears a red flag
| When you're open | When you're closed | By text, WhatsApp or chat | |
|---|---|---|---|
| Immediately | Tilly says: "What you're describing could be an emergency. I'm connecting you to our team right now — please stay on the line." It rings everyone on your urgent line at once, then your front desk. | Tilly reads out your emergency partner's name, number and address, slowly. | Within seconds, Tilly replies asking the client to call you now, with your emergency partner's details. |
| While ringing | Your staff hear a short summary ("Urgent: possible toxin, dog, Biscuit") and press 1 to take the call. The caller hears a short reassurance every ten seconds. | Tilly texts the details to the caller and offers to connect them to the emergency clinic. | Your team gets an alert with a one-tap "Call client" button. |
| If no one answers | After 30 seconds (you can set 15–60), Tilly gives the emergency partner's details by voice and text and alerts your on-call vet. | Your on-call vet gets a phone call, a text and an app alert. If they don't acknowledge within 5 minutes, Tilly alerts the next person on your rota, for up to three rounds, then texts owners and managers and pages Tilly support. | After hours, the on-call alert runs the same way. |
| In your dashboard | A red banner appears for everyone at that location within two seconds and stays until someone acknowledges it. | Same. | Same. |
| Afterwards | An urgent follow-up task, due within 60 minutes (or 60 minutes after you next open). | Same. | Same. |
| Review | Your clinic and a Tilly clinical advisor review every urgent conversation the same day (by noon the next day if it came in after 8 p.m.). | Same. | Same. |
If the caller hangs up early, Tilly still texts the emergency partner's details, alerts your on-call vet after hours and creates the follow-up task. If the caller's number can't receive texts, Tilly repeats the number slowly, twice. During an emergency, Tilly only speaks fixed lines approved by veterinarians, recorded in advance in your clinic's Tilly voice, so a technical problem with live speech can't stop them being heard.
Timings are Tilly defaults. Some can be adjusted by your clinic within safe limits.
Some calls need a person's compassion
Not every hard call is an emergency. When a caller wants to talk about the end of a pet's life, a pet that has died, or something that went wrong, they need kindness and a person, not a booking flow. Tilly recognises these calls and passes them to your team with care.
Calls Tilly hands to your team
- Euthanasia and end-of-life questions
- The death of a pet, and grief
- Complaints about care
- Questions about how a surgery went
- Billing disputes
- Callers who are abusive
What Tilly does
- Opens with fixed, kind wording approved by veterinarians, then keeps to short, calm sentences
- Stops whatever it was doing and doesn't push a booking, a reminder or a sale
- Offers to put the caller through to a person when someone is available, or a priority callback
- After hours, gives a caller asking about euthanasia your emergency partner's number in case their pet is suffering now
- Tells your team who called and why, so nobody has to ask twice
- Never gives clinical details, argues about a bill or promises a refund
- Never books a euthanasia, whatever your settings
Example from our demo clinic
- Caller
- Our cat Milo died this morning and I don't know what to do next.
- Tilly
- I'm so sorry for your loss. Thank you for letting us know. I'll let the team know, and make sure we stop sending reminders for Milo. I can put you through to someone on the team now, or ask them to call you back as a priority — whichever is easier for you.
Your clinic chooses who takes each kind of sensitive call and in what order; it can't switch off the protocol, change its wording or remove the offer of a person or a priority callback. A possible emergency always comes first. Sensitive calls appear in your dashboard marked "Sensitive".
Rules written by vets, signed off by yours
Written by licensed veterinarians.
Tilly's clinical advisors write the red-flag list and the exact words Tilly uses in an emergency, in every language.
Tested before use.
Each new version of the rules is run against our safety test set of simulated calls in every language. It can only be published if it catches every emergency in the set. Native speakers and a clinical advisor approve each language.
Signed off by your vet.
Before your clinic goes live, a licensed vet at your practice reviews the rules, listens to sample emergency calls in your own Tilly voice and signs them off with their name and licence number.
Your additions, never subtractions.
Your vets can add red flags for your clinic, for example a breed you see often. Nobody at your clinic can remove a base red flag or make one weaker.
Every urgent call reviewed.
Your team and a Tilly clinical advisor licensed in the US review each urgent conversation the same day.
Learning from near misses.
If anyone at your clinic flags a conversation that should have been urgent, it goes to our clinical advisors and becomes a new test case.
What Tilly will never do
Tilly will never:
- Give medical advice, a diagnosis, a medicine name or a dose.
- Suggest home remedies or first aid.
- Tell a caller that something isn't serious.
- Lower the urgency of a call once it's flagged.
- Look at, describe or judge a photo.
- Read out or interpret test results.
- Pretend to be a person or use a human name as its own.
- Refuse to connect a caller who asks for a person.
- Take card numbers over the phone or by message.
- Dial 911 or any public emergency number itself.
- Contact a client for reminders or marketing without their recorded consent.
- Leave a caller in silence: if anything fails, the call rings through to your clinic.
- Book a euthanasia, or push a booking on a caller who is grieving or upset.
Callers always know who they're talking to
"Hi, I'm Tilly, the AI assistant at Maple Street Animal Hospital."
Every call and every new message thread starts like this, in the caller's language. You can rename Tilly, but the name can't be a person's name from your team, can't include titles like "Dr" or "Vet", and the words "AI assistant" always stay. If a caller asks "Am I talking to a real person?", Tilly says it's an AI and offers to connect them to your team.
If your clinic records calls, callers hear "This call is recorded so the clinic can review it. If you'd prefer not to be recorded, just say so." If they object, recording stops at once and the audio captured so far is deleted. The call carries on as normal. You can also choose not to record at all.
Clear lines of responsibility
| Tilly | Your clinic |
|---|---|
| Recognising possible emergencies and routing callers to people | Clinical decisions and advice |
| Keeping the base red-flag rules and emergency wording up to date, approved by licensed vets | Signing off the rules before go-live, and approving any red flags you add |
| Making sure urgent calls ring your urgent line, emergency partner and on-call rota | Keeping those numbers, hours and rota correct |
| Logging and reviewing every urgent conversation | Reviewing and acknowledging urgent conversations in your dashboard |
Questions
Is it legal for an AI to answer a vet clinic's phone?
Answering calls, booking and taking messages is front-desk work. Giving veterinary advice is not, and Tilly doesn't do it. Tilly tells every caller it's an AI, as US state laws expect.
What if Tilly misses an emergency?
The system is designed to over-call, not under-call: any doubt counts as urgent. Every rule change must catch every emergency in our safety test set before use. If your team spots a conversation that should have been urgent, one click sends it to our clinical advisors.
Can the AI model override the safety rules?
No. Emergency detection runs before the model on every sentence, the urgent path doesn't use the model at all, and a separate check reviews everything Tilly says before it's spoken.
What does Tilly do with photos clients send?
It attaches them for your team and tells the client it can't look at photos. It never comments on them.
Does Tilly work in other languages safely?
Yes. Red flags and emergency wording exist in every supported language and are approved by a clinical advisor and native speakers before use.
Who reviews urgent calls on weekends?
Tilly clinical advisors review urgent conversations on the same schedule every day. Your clinic reviews them when you're next in, and the urgent banner stays until someone acknowledges it.
Can Tilly book a euthanasia appointment?
No, never. Calls about euthanasia, the loss of a pet, complaints and other sensitive topics go to your team: Tilly answers with fixed, gentle wording approved by veterinarians, offers a person when someone on your team is available or a priority callback, and doesn't try to book anything.
Hear the emergency path for yourself
Call our demo and say "show me an emergency."
Book a walkthrough: with a member of our team, including our clinical approach