Ask Wodify

From a chat box to an AI experience in one week

Ask Wodify is an AI assistant that lets gym owners ask questions about their business in plain language. Engineering had a working backend and a basic chat window. In a one-week design sprint, I turned that starting point into a cohesive, extensible product experience, using AI-assisted prototyping to move from concept to working interactions.

ROLE

Product Designer

FOCUS

Conversational AI, Interaction Design, AI-Assisted Prototyping

TIMELINE

~1-week design sprint

STATUS

Launched

01 · Starting point

Starting with real customer workflows

Ask Wodify didn’t start from a feature list. Some of our most AI-curious customers were already building their own workflows on top of their Wodify data. Their use cases showed the team where an integrated assistant could help.

By the time I joined, engineering had a functional backend and a bare-bones chat interface. My job was to turn it into an experience that felt native to Wodify and could grow with future tasks, in about a week.

01

Customer workflows

AI-enthusiast customers were already building their own workflows with Wodify data.

02

Engineering prototype

A working backend and a basic chat interface.

03

Design challenge

Shape a cohesive, extensible experience in one week.

02 · Interface

Designing beyond a chat box

Layout

I prototyped two directions with Claude Code: a partial-screen panel beside existing pages, and a full-page experience. Working prototypes let stakeholders compare them in context. We chose full page because it leaves room for richer responses now and more complex tasks later.

Familiar patterns

Rather than invent a new surface, I adapted patterns from Wodify’s existing Inbox, so the assistant feels like part of the product owners already use.

Interaction foundation

Three patterns carry the experience: suggested prompts drawn from real customer use cases, a reusable structured question with selectable options for when the assistant needs clarification, and the messaging interactions around them.

PLACEHOLDER

Partial-screen layout

Add a capture of the partial-screen prototype.

PLACEHOLDER

Full-page layout (chosen)

Add a capture of the full-page prototype.

PLACEHOLDER

Annotated interface

Add the final UI with callouts for suggested prompts, structured questions with selectable options, and messaging elements.

03 · Process

Using AI to design an AI product

A one-week sprint left no room for a long handoff, so I built AI into how I worked while keeping the design decisions mine.

Prototyping

Claude Code let me explore UI directions as interactive prototypes, so reviews focused on how the experience behaved, not only how it looked.

Writing

A custom UX writing skill helped me refine interface copy, from suggested prompts to system messages.

Decisions

A custom meeting-documentation skill captured design decisions as we made them, so the reasoning was easy to find later.

Handoff

I kept Figma artifacts alongside the prototypes. Engineering used the prototypes as a reference to rebuild the production UI.

01

Explore

Interactive prototypes built with Claude Code.

02

Refine

Interface copy shaped with a UX writing skill.

03

Record

Decisions captured with a meeting-documentation skill.

04

Hand off

Prototypes and Figma as references for engineering’s rebuild.

Designing for the edges

I also designed the moments around the conversation: a loading animation that shows the assistant is working, and states for when a conversation reaches its token limit.

PLACEHOLDER

Loading animation

Add a recording or GIF of the loading animation.

PLACEHOLDER

Token-limit state

Add the token-limit state design.

04 · Outcome & reflection

Shipped, with a foundation to grow on

Ask Wodify launched as a product. The sprint set its interface foundation: a full-page layout, a set of reusable conversational patterns, and considered states for loading and token limits.

Reflection

Working prototypes changed what I think of as a design deliverable. Instead of describing interactions in static frames, I could show them, test them with stakeholders, and give engineering something concrete to build from. The question I’m still exploring is which decisions belong in a prototype and which still belong in Figma.

WAIT! THERE’S MORE

There’s more to this story than I can share here.

Due to confidentiality, this is only a glimpse. I’d love to walk you through the prototypes, the layout explorations, and the interaction details in more depth.

Prototype walkthrough

Layout explorations

Interaction details

Design decisions

@2026 Sanni Zhang

Open to new positions. Let’s work together!