K3 Labs is a web3 workflow automation platform that combines conversational AI with visual workflow building to simplify blockchain interactions.

The Challenge

K3 Labs' drag-and-drop builder felt dated as users increasingly expected conversational, AI-driven interactions. The challenge was to reimagine the experience as chat-first while preserving the precision and control of visual workflow building.


Product Goal

Create a new product experience that leads with conversational AI while thoughtfully integrating visual workflow tools, allowing users to build automation through natural language while maintaining access to advanced capabilities when needed.


My Impact

  • Sole designer collaborating with product leadership to transform the product from workflow-builder-first to AI-chat-first

  • Shipped 20+ user flows and high-fidelity designs for December MVP launch

  • $5M funding goal: MVP designed to demonstrate product vision to investors

  • New AI capabilities: designed plug-and-play MCP connections and integration features

  • Design system: built from scratch to scale across multiple brands

Role
Sole Product Designer


Deliverables
20+ user flows with high-fidelity mockups for landing page, dashboard, and chat-first experience


Timeframe
Sept-Dec 2024, 3 months

Competitive Analysis

I analyzed 20+ competitors in the chat-first and workflow builder space, organizing findings into 7 themes 
such as AI-native chat experiences, pre-built prompts, registration patterns, and initial workspace interactions. 
The competitive landscape revealed that no platform fully integrates conversational workflow building 
with comprehensive visual workflow editing, an opportunity our solution addresses.

Usability Audit: Workflow builder

I conducted a usability audit of the current workflow builder, informed by competitive analysis of node-based builders and chat-based workflow tools.

  • Incomplete workflows go unnoticed:
    Validation feedback missing when required fields are incomplete

  • Technical barrier for non-coders:
    Nested menu structure requires coding knowledge

  • No conversational interface:
    Users can't build workflows through natural language

  • MCP integration unclear:
    Process for connecting Model Context Protocols is vague

  • Blind workflow building:
    No real-time preview of workflow execution

Interlinkeable visual language

To bridge the old workflow system with the new node-based builder, I developed an interlinked visual language. Color-coded tags and breadcrumbs appear in real-time within the chatbot, while corresponding node iconography appears in the workspace to the right, creating a cohesive experience as users build workflows.

Scalable Design System

K3 Labs' Web3 operating system is designed to be adopted by multiple brands as their AI chat-based interface. Starting with no design system in place, I created a scalable token architecture that prioritizes both developer efficiency and brand flexibility.

User Flow

Successfully launching a Coinbase Trading Agent

Previously, users had to sign up before deploying agents, creating drop-off. I redesigned the flow to enable instant deployment through the homepage chat interface. This eliminates friction at the critical first touchpoint while demonstrating the chat-first experience that differentiates the platform.

Click image to enlarge ↑

Information Architecture

I mapped the product's conversational architecture through 20+ user flows, defining how the AI responds to different user inputs and guiding users through complex tasks via chat. Weekly iterations with stakeholders and the Product Lead helped me refine these interaction patterns into high-fidelity designs that balanced AI capabilities with user expectations.

Next steps

With launch planned for the coming weeks, the next phase focuses on learning and iteration. Early user feedback from investors and partners will inform prioritization of refinements. As a new product, continuous refinement based on real user needs will be essential to achieving product-market fit.

Next Project Link
Next project: Funnelish →