The problem
A neighborhood veterinary clinic was losing appointments on WhatsApp: messages came in at all hours and were only answered the following day. But the real problem was different. Clients were used to the receptionist they knew, to the way she replied — and a generic bot gives itself away in the first message.
What we did
Before writing a single line of code we studied how she worked: three months of real conversations, read to understand how she greeted people, how formal she was with whom, how she asked for the pet’s details, how she escalated an emergency, and when she wrapped up with a “see you tomorrow”. That produced the two pieces that make the difference: the assistant’s personality and a set of real examples it uses as reference when replying.
On that base we built the system: it books appointments against the clinic’s real Google Calendar, reviews the reason before confirming, and routes emergencies to a person. The owner adjusts that personality, adds examples, and updates business information from a dashboard, without touching code. When a conversation gets complicated the team takes it over by hand and the assistant goes quiet in that chat. We also added a booking portal for people who would rather not write.
The outcome
Live and answering every day. What we like most about this case: clients write to it the way they used to write to her, and almost nobody asks whether they are talking to a bot.
Built with
- Google Calendar
- Claude
- Custom admin dashboard
This case is part of AI Assistants
See the service
