AI-Powered Design Delivery
For design team internal shareout
Xiaoye Lin
May 2026
Applying AI in Our Current Design Process
A Real Project at TikTok Shop Seller Affiliate · Q2 2026
EU Seller Targeted Invitation
Project Background

When creating targeted invitations in Affiliate, seller can select preferred countries when they create target collaborations. The PM prepared the PRD and kicked off the project with me.

Using AI in Early-Stage
Design Exploration
For better understanding and more efficient communications
Tools
How AI Can Help — #1
A Working Prorotype Reveals Hidden Blind Spots Early
Figma version
Demo in TRAE Fix
How AI Can Help — #2
More Efficient Communication with Cross-functional parterners
Talk to PM
How AI Can Help — #3
Efficiently Look Up & Verify Information
Tools Prototype Desk Research Ask PRD Verify
Pitfall
UI Fidelity Issues
1.Figma plugin
Figma Plugin
Production
Production
Before Plugin
Without Plugin
After Plugin
With Plugin
2.Inspire Prototype Comes with the TikTok's Design System
Inspire Prototype
3.Hand-coded Design System Based on Working Experience
GitHub
4. Ask the Dev Team for a Real Development Environment
Dev Environment

Every project is different — plan early, understand the design requirements, and choose the right AI capabilities accordingly.

Using AI After
Design Handoff
Solving the team's pain point: gaps between dev implementation and design specs. Designers step in as fine-tuners.

Once dev notified and aligned on the changes, my fine-tuning work began...

RD RD Doc
Vibe Coding Fine-Tuning in Progress
Pushing Code to Dev
Design Done
Dev Done
Thanks to Jingyi and the dev team for enabling the "local service install" capability
ContentX
How AI Helped — #1
Design-to-Dev Implementation Accuracy Significantly Improved
Design Spec
Design
Dev Result
Dev
After Vibe Coding
Tweaked
Design Spec
Design
Dev Result
Dev
After Vibe Coding
How AI Helped — #2
Fixed some issues directly — no need to wait for dev, boosting team efficiency
Small Tweak 1 Tweak Request
Small Tweak 2 Tweak Info
Pitfall #1
Technical Barriers
6+ Hours Debugging in the Darkness
Test
Unpredictable Issues
Unexpected Issue
Pitfall #2
Modified Code Still Needs Dev Cleanup
1. Fine-tuning UI still heavily relies on components in the dev environment. AI may modify them on its own, which is unpredictable and may bring unneccessary impact.
2. AI-generated codes may not meet requirements, so it still needs manual refinement from the dev team.
Asset Issue
With the Power of AI

AI vibe coding tools unlock designer's power, blurring the boundaries of roles and increasing influence of design function in an org.

Design Core: Solving Human Problems.

We still apply empathy, independent thinking, and critical judgment to be successful.

Illustration

illustration by ChatGPT

← → Navigate