Heuristic evaluation
Flags usability risks using established interaction principles and severity levels.
UX-Ray. AI-Assisted UX Evaluation for Scalable Design Workflows
An independent R&D project exploring how AI can reduce audit friction and bring clearer feedback into Figma.
Product website
Demo video here, or explore the live UX-Ray site for the full product story.
"The goal was never to replace designers. The goal was to reduce operational friction, accelerate evaluation cycles, and bring clarity to design decisions through AI-assisted workflows."
Executive summary
A UX audit layer inside the design tool
UX-Ray explores how AI can support design evaluation directly in the environment where designers work. The plugin analyzes Figma frames against usability heuristics, accessibility indicators, visual hierarchy principles, and custom brand guidelines.
The project responds to a practical operations problem: UX reviews often happen late, vary by reviewer, and consume time that product teams rarely have. UX-Ray was designed to provide faster, structured feedback while keeping strategic judgment in human hands.
The problem
Quality review does not scale naturally with product complexity
As products and design systems grow, evaluation work becomes harder to run consistently. Teams repeat accessibility checks, visual reviews, content critiques, and heuristic analysis across many screens and iterations.
The friction is not only time. It is uneven quality, late feedback, missed consistency issues, and cognitive load for designers who need to make decisions quickly. UX-Ray started with a question: how can AI help teams scale evaluation without reducing the role of human designers?
Strategic vision
AI should support UX workflows, not replace human design thinking. UX-Ray acts as an operational assistant that surfaces patterns, explains risks, and helps designers focus their attention.
Core capabilities
Multiple evaluation layers in one review pass
Accessibility validation
Reviews contrast, readability, and WCAG-style indicators to surface common barriers.
Visual hierarchy analysis
Identifies spacing, alignment, structure, and emphasis issues that affect scanning.
Brand consistency review
Compares interface choices against custom typography, color, spacing, and style rules.
Operational challenge
Repeatable review standards without rigid rules
- Repeated checks: accessibility, hierarchy, labeling, and brand drift across many screens.
- Bandwidth: review quality depended on who was available.
- Direction: accelerate discovery of issues worth discussing while designers keep judgment.
Technical architecture
From Figma frame to categorized recommendations
- Extract design structure and semantic metadata from selected Figma frames.
- Generate visual and contextual analysis inputs for the evaluation prompt.
- Send structured prompts to GPT-4 Vision for review.
- Parse AI feedback into categories, severity levels, and explanations.
- Return actionable recommendations directly inside the design workspace.
Enterprise potential
From plugin to design operations layer
The longer-term direction: team-wide evaluation, organization-specific guidelines, and AI-assisted governance for distributed or regulated teams—focused on consistency and review acceleration.
My role
- Product strategy and plugin UX
- AI workflow and prompt structures
- Technical integration concept for in-Figma feedback
Outcomes
A working concept for faster feedback loops, more consistent UX reviews, and less repetitive operational work—with a path toward scalable design operations.
Reflection
The insight was not AI replacing expertise—it was using structured analysis to improve clarity, consistency, and operational flow in product organizations.
UX-Ray is an independent R&D concept and is not affiliated with Figma or OpenAI. Performance figures and projected improvements are based on internal experimentation, industry benchmarks, and exploratory testing rather than production-scale deployment.