I owned product design end to end, from turning the founders’ ideas into a scoped MVP to reviewing builds before release. I built the design system by extending Shadcn’s primitives with our brand tokens and custom components, so the team could design and build from the same parts. When the team expanded beyond the chatbot, I scoped and designed the website builder that followed.

0→1

Design Systems

MVP Scoping

0→1

Design Systems

MVP Scoping

This page shows the shipped product and the design decisions behind it. The research stays with NEXT,

so think of it as my resume bullets for this role, with visuals.

This page shows the shipped product and the design decisions behind it. The research stays with NEXT, so think of it as my resume bullets for this role, with visuals.

Starting without a blank page

Train your AI” is developer language. For a business owner, it’s an empty box with no obvious starting point. So I gave them three ways in: paste a URL and the platform fetches the content, write plain-text instructions, or start from Bot Introduction, a generated baseline. Editing something is easier than writing from nothing.

Three ways to start: fetch from a URL, write plain instructions, or edit a generated baseline.

See it before your customers do

Owners needed to trust the bot before putting it on their site. Every change in the edit panel updates the chatbot instantly, in both its closed bubble and its open window. Owners also control the conversation itself: the first message, and whether to collect an email or phone number. Then “Try your chatbot” lets them talk to their bot exactly as a customer would, and fix what’s off before going live.

Every edit shows up instantly, and owners test the bot as their customers will.

From chatbot to full digital presence

Once the chatbot was live, owners started asking for a website too. So the builder opens with a site already generated from the material they uploaded to train their chatbot, with that chatbot embedded and working from day one. Owners edit something real instead of starting over.

Making the builder simple for our existing users took the most iteration. I ran usability testing and revised the design over several rounds with the developers. Owners work in prebuilt blocks: add, remove, or reorder sections, edit text inline, and see in a side panel exactly what each element can change. Advanced options like Google Analytics and SEO stay in More settings, available to owners who want them and out of the way for everyone else.

The site arrives already built from their chatbot material, with the chatbot embedded from day one.

Outcome & Reflection

The chatbot builder found early traction, especially within the Chinese small business community. In the three months after the website builder launched, NEXT’s user base grew 70%.

Being the first designer meant designing the process as much as the product. I learned to build feedback into my own workflow: walking developers through flows before handoff, testing with real owners early, and treating the design system as the team’s shared language rather than a Figma file. I want to bring that ownership to a team where it meets critique from other designers.