I Fed Our Company’s UX Design Guideline to an AI — and It Learned to Make Design Judgments

1. Design’s essence is a ‘decision tree’

Design is not just about intuition, but a sequence of judgments filtered by conditions and rules (a decision tree).

AI is particularly strong at this kind of rule-based reasoning and decision-tree processing.

2. Turning guidelines into an ‘LLM wiki’

The author trained an LLM on many of the company’s UX design guideline documents and turned them into a conversational ‘LLM wiki’.

As a result, it went beyond simply searching for rules and began independently inferring and presenting the design elements and reasons suited to specific situations (for example, how to handle notifications when deleting emails in bulk).

3. Practical application and results

The design system components were rebuilt on a general HTML-based foundation and connected to an AI agent.

With only requirements and screenshots, it generated HTML prototypes that were immediately usable at a 70–80% level, reducing the time needed for first drafts from the usual 1–2 weeks to 1–2 days.

4. The designer’s role in the AI era

While repetitive, rule-based layout decisions can be replaced by AI, designers will focus on higher-level work such as the following.

Defining the right problem: the ability to reframe the problem itself (for example, implementing autosave instead of adding a notification dialog)
Creating new interactions: designing new interfaces/experiences beyond existing rules
Team persuasion and politics: aligning with stakeholders and carrying the responsibility for decisions
True user empathy: capturing subtle user reactions in context
Wiki and system maintenance: advancing the structure of guidelines and keeping them maintained

XPLORED View

By letting AI trained on design rules handle repetitive judgments and prototype creation, designers can move beyond routine tasks and step into their essential roles: problem definition, creativity, and empathy.

Good timing to read this

When there are many design guidelines, but the actual decision criteria remain implicit knowledge within the team
When you want to connect AI to the design system, component selection, and prototyping workflow

What to take away

Design can be seen not just as intuition, but as a decision tree in which conditions and rules are linked together.
As repetitive judgment becomes automated, the designer’s role shifts toward problem definition, creativity, empathy, and persuasion.

1. Design’s essence is a ‘decision tree’

2. Turning guidelines into an ‘LLM wiki’

3. Practical application and results

4. The designer’s role in the AI era

XPLORED View

Get the week's signal by email, every Saturday.