Introducing In-app Support

Overview

Garden's support lived on Discord. To get help, a user had to figure out support was on Discord, leave the product, join the server, find the right channel, and post their question. Most users didn't bother; they gave up silently. So I designed and shipped an in-product support system: an AI assistant trained on Garden's docs, app state, and on-chain explorer data, with human handoff to the team when needed.

Daily ticket volume doubled within weeks, not because users had more problems, but because they finally had a direct line.

My Approach

Discord was working as a forum, not as a support channel; it filtered out everyone who didn't already know to look there. We had to design the support where the problem happens.

What It Does

• Captures every interaction in a structured log, replacing scattered Discord tickets. Patterns across user issues are now visible at the macro level.

• Lives inside the product. Users get help in the same window they're trying to swap in; no platform-switching.

• Answers basic queries automatically using docs, app state, and on-chain explorer context. Cuts the team's response load on routine questions.

• Routes feedback back into product. The team sees not just what users ask, but also the gaps in the product surface as questions.

Why It Matters

Doubling support volume is usually a bad signal. Here it was the win. The in-app system captures their voice in real time and turns it into a feedback loop the product team uses. Support stopped being a place where problems die and became one of the highest-signal data streams Garden has.