Specialist clinic — Healthcare
An assistant that answers before the front desk can
Most incoming messages were the same few questions about appointments and pricing. We built an assistant that answers them from the clinic’s real information.
- Year
- 2025
- Industry
- Healthcare
01Challenge
Hundreds of messages arrived daily across Instagram and WhatsApp and the front desk could not keep up. A large share were the same handful of questions: when is a slot free, what does this treatment cost, where are you. The genuinely important messages got lost among them.
02Analysis
- 01Categorising the real messagesWe classified a thousand past messages: close to seventy percent fell into five fixed categories.
- 02Where a human is requiredClinical questions and sensitive cases must never go to an assistant. We drew that boundary before building anything.
- 03The cost of a late replyMessages left unanswered for more than two hours were noticeably less likely to turn into a booking.
03Solution architecture
01 — Solution architecture
Channels
- Site form
02 — Solution architecture
Assistant
- Message classification
- Answers from clinic knowledge
- Escalation boundary
03 — Solution architecture
Knowledge
- Appointment calendar
- Treatments and pricing
- FAQ
04 — Solution architecture
Human
- Front-desk queue
- Sensitive cases
- Sample review
04Results
- Time to first replyup to several hoursunder a minute
- Handled without a person0a meaningful sharerepetitive, non-sensitive questions
- Escalated to the front deskeverythingonly cases needing judgement
ToolsClaude APIRAGn8nBooking calendar API
Next step
Does your project look like this?
If any part of that story felt familiar, we can probably help.