AI chatbot cuts support tickets by 28% for a wellness brand

A large yoga and wellness brand had thousands of pages of content and a marketing team answering the same questions all day. We built a chatbot that gives the exact right answer, not a nearly right one.

The problem

The business spans a hotel, a café, day classes, retreats and teacher training across Australia and Thailand. Its website had 2,000 to 3,000 pages covering dozens of yoga styles that differ in small but important ways, plus a community forum built up over five to ten years.

Visitors couldn’t find what they needed. The small marketing team was stuck answering questions about pricing and discounts instead of marketing, and sales slowed down.

What we built

1. 01

   A chatbot that understands the whole offer

   It guides each visitor to the right information, whether they’re booking a single class or training to become a teacher. When a question needs a person, it hands over smoothly.

2. 02

   Precise, not just close

   A standard AI fed the website gave answers that were nearly right. But when someone asks which discount applies to which course, nearly right isn’t good enough. After real trial and error, we combined vector search with a graph database (Graphiti), so the system understands how programmes and prices relate to each other.

3. 03

   Years of forum answers, without the outdated ones

   The forum went in too. Because prices and programmes changed over the years, the same question often had different answers. The system learned to always give the current, correct one.

4. 04

   Every conversation in HubSpot

   Each chat and lead is logged in the CRM automatically, and the bot mentions the right offer when a visitor shows interest.

The result