How Coffee.ai turned a disorganized help center into a support asset that actually works

The situation

Coffee.ai had a help center. The problem was it wasn’t helping anyone. Content was scattered, the structure reflected how the team thought about the product — not how users experienced it — and Intercom’s Fin AI agent was returning poor responses because the underlying articles weren’t written with AI retrieval in mind. Users were landing in the help center and still raising tickets because they couldn’t find what they needed.

What we did

We started with an audit of the existing help center, cross-referenced against their support conversations to understand what users were actually asking. What we found was a structure organized around features rather than user goals. We rebuilt the information architecture from the ground up around user intent, rewrote every article in plain language optimised for both search and Fin AI retrieval, and reconfigured the Fin agent so it could surface the right content reliably.

We also took on their ongoing monthly changelog — translating each product release into clear, user-facing language that the team no longer had to think about.

The result

Coffee.ai now has a help center that works as a genuine support asset — content users can find, articles Fin can use, and a changelog that keeps users informed with every release.

“Tovana is very knowledgeable about Intercom and how to structure support content. They audited and re-drafted our entire help center and improved Intercom’s Fin responses immensely.”

— Doug, CEO — Coffee.ai

‍