ChatKit
Designing an AI chatbot builder for teams who don’t write code.
Non-technical teams locked out of their own chatbots.
AI drafts the flow. The user shapes it.
Zero engineering tickets for content updates.
The Problem
Non-technical teams were locked out.
The existing chatbot platform was built for developers. Every change — a new response, an updated flow, a corrected answer — required an engineering ticket. Support teams had the ideas. They just had no way to act on them.
The backlog grew. The bots stagnated. The people who knew the customers best had no seat at the table.
Guided bot setup — define purpose, tone, and scope before touching the builder.
The Insight
The blank canvas was the problem.
Sitting with support leads trying to use the old builder, the issue wasn’t ability — it was confidence. A blank canvas implied expertise they didn’t have. So we flipped the model: instead of building from scratch, users start with an AI-generated draft and edit from there.
From builder to editor. The same outcome, half the fear.
The conversation builder — AI drafts the flow, the user shapes it.
Design Decisions
Three principles that held.
Automation vs. control. AI handles the tedious scaffolding. The user confirms, edits, and steers. The system should feel like a capable collaborator, not an autocomplete.
Guided, not gated. Onboarding as first principle — define purpose, tone, and scope before touching the builder. Structure up front means fewer wrong turns later.
Knowledge as a first-class surface. The knowledge base isn’t a settings page. It’s core to the product — transparent, editable, and directly connected to what the bot says.
Testing and knowledge — validate before going live.
0
Engineering tickets needed for content updates
↑
Confidence among non-technical support teams
↓
Time from idea to deployed change
“The best outcome wasn’t that the tool was easier. It was that people stopped assuming they couldn’t do it.”