When Design Systems Meet AI: Shifting the FE Development Paradigm
디자인시스템이 AI를 만났을 때: FE 개발 패러다임의 변화
HCI Today summarized the key points
- •This article is based on a presentation at NAVER ENGINEERING DAY 2025, where the speaker shares their experience automating markup with a design system and AI.
- •First, the presenter explains the preparation process by reviewing design tokens and components in NAVER Financial’s design system.
- •Next, they discuss how AI can be used to help with markup work, and how they prepared the necessary reference materials in advance.
- •After applying it in practice, they said some parts were good enough to use in development, but there were still areas that left something to be desired.
- •Rather than fully replacing markup work, this article highlights AI’s potential as a tool for reducing repetitive tasks.
This summary was generated by an AI editor based on HCI expert perspectives.
Why Read This from an HCI Perspective
This article is meaningful from an HCI perspective because it doesn’t treat AI as merely a ‘smart tool,’ but shows how it can be used and refined within real work workflows. Elements such as design systems, markup automation, and upfront preparation clearly demonstrate that success is driven less by raw technical performance and more by user experience and the structure of work processes. For practitioners, it offers practical hints they can try right away; for researchers, it raises questions about how people and AI collaborate.
CIT's Commentary
The core of this case is that AI doesn’t simply ‘do the markup for you’; rather, it depends on what preparation and constraints are in place to reduce the team’s workload. The better the design tokens and components are organized, the easier automation becomes. But if the structure is ambiguous, AI may produce plausible output while still requiring more manual effort during actual development. So what matters is not model performance alone, but the design of interfaces and rules. In particular, markup automation must be designed not only around the final output, but also around intermediate review, the path for edits, and what happens when things fail—how to roll back. This viewpoint holds true not only for large organizations’ design systems like NAVER’s, but also for fast-moving environments such as domestic startups. Ultimately, adopting AI should be read as a question of how safely and predictably teams can collaborate—not simply how much automation is achieved.
Questions to Consider While Reading
- Q.In markup automation, where does AI fail most often: understanding structure, component mapping, or exception handling?
- Q.I’d like to know how you determined where humans must step in, and how those criteria affected both work time and quality.
- Q.How would the outcomes differ—even with the same AI tool—between organizations with a well-established design system and those without?
This commentary was generated by an AI editor based on HCI expert perspectives.
Please refer to the original for accurate details.
Subscribe to Newsletter
Get the weekly HCI highlights delivered to your inbox every Friday.