Case study

DLS Magician

Designing an AI-assisted workflow that turns ambiguous text requirements into explainable, editable, DLS-based prototypes.

ACM IUI Paper US Patent
Problem
E&P product teams relied on text-heavy requirements, while product owners needed visual prototypes earlier to align stakeholders and validate direction.
Role
Designed the AI-assisted workflow connecting product-owner user stories, natural-language parsing, DLS component logic, and generated prototype output. A core design principle was making the AI's reasoning visible at every step — so teams could inspect, question, and refine the output rather than accept it blindly.
Design decision
Designed a structured translation layer between natural-language requirements and DLS components, making generated prototypes explainable, editable, and aligned with platform conventions.
Research
Reviewed 30 internal project records, surveyed product owners, and interviewed cross-functional participants across the US, Germany, UK, and Malaysia.
Impact

Reduced manual prototyping effort by 45%, enabling teams to move from text requirements to reviewable prototypes in minutes instead of weeks.

Made AI-generated UI output easier to inspect, discuss, and refine before engineering investment.

01

Requirement signal

User stories became structured input that product owners could inspect and refine.

02

AI translation layer

NLP parsing connected ambiguous text to product intent, interaction needs, and UI structure.

03

DLS component mapping

The system matched requirements to reusable design language components and patterns.

04

Prototype review

Generated prototypes made AI output visible, editable, and easier to discuss before build.