How Golf New Zealand built Kim, an AI agent serving 10,000+ member conversations across New Zealand

At a glance

Golf New Zealand is the governing body for golf in the country, responsible for the official handicap system, the national representative programme, and more than 20 elite tournaments including the New Zealand Open. With a nearly 100% self-serve membership base relying on the website and app, support volume was high and growing. Jon Ward, Digital Experience Manager, needed a smarter way to handle it.

A part-time inbox that was always overflowing

Before Chatbase, Golf New Zealand had a dedicated part-time staff member whose entire role was responding to member queries by email. It was manual, slow, and expensive, and it never quite worked. The inbox would overflow onto other team members, pulling them away from their own responsibilities. Members were also telling Golf New Zealand that the website was hard to navigate. They wanted a search bar. They wanted faster answers. The cost of the status quo was both financial and operational, and it was becoming impossible to ignore.

Meet Kim, Golf New Zealand's always-on member assistant

Golf New Zealand deployed their Chatbase AI agent on the website and named her Kim. Kim answers member queries directly in the chat, helps people find resources on what Jon describes as a "vast website", and troubleshoots common issues like forgotten passwords. She supersedes the need for a search bar entirely: members just ask her where to find things. Kim was trained on a broad set of sources: web links, resource articles, video transcripts, and a growing library of manual training notes added over time as new patterns emerged. Setup followed a YouTube tutorial and was, in Jon's words, "pretty simple and straightforward."

What they built

• AI agent embedded on the Golf New Zealand website, handling member queries 24/7
• Trained on web links, resource articles, video transcripts, and manual training notes
• Running in tandem with a HubSpot AI email agent for inbox queries, both named Kim
• Ongoing training loop: thumbs-down conversations reviewed regularly and used to improve responses

The results

Kim has handled over 10,000 conversations and 43,000 messages, and the team is continually surprised by how well she has been received. 169 thumbs up against 111 thumbs down, a positive ratio that has held as volume has grown. The manual inbox workload has been dramatically reduced, and the team's time is no longer being absorbed by routine queries. Members get faster answers, and the website feels more navigable, without a single line of additional development work.

What's next

Golf New Zealand is exploring CRM integration to give Kim a fuller picture of member data, and wants to build out custom actions and widgets to improve the agent's ability to take real steps, not just answer questions. The foundation is solid: Kim has proven the model, and the team now wants to push it further.