Agentic UX Workflow

MY ROLE

Concept Design & Dev

COMPANY

1und1 / GMX Media

TIMELINE

March -April 2026

PROJECT DESCRIPTION

Lumen explores how agentic AI workflows can be used to design complex, content‑driven products faster and more coherently—without sacrificing product intent, editorial trust, or UX quality. The project focuses on orchestrating AI tools through clear constraints rather than using them as generative shortcuts.

APPROACH: AGENTIC DESIGN WORKFLOW

Instead of treating AI as a single tool, Lumen was designed using an agentic workflow—where AI systems operate within clearly defined roles, inputs, and constraints.

The workflow combined:

  • a structured Product Requirements Document (PRD)

  • an existing design system

  • and Claude Design as a reasoning and synthesis layer

This allowed design decisions to be explored, challenged, and refined conversationally.

HMW QUESTION

How might we help young people stay informed in a way that fits their media habits while still giving them the context and clarity traditional news often lacks?

PRIMARY RESEARCH QUESTION

Primary research combined secondary data, a user needs framework (know, understand, feel, act), and interviews supported by Copilot to structure insights and refine questions. This setup ensured that findings translated directly into product requirements rather than abstract observations.

INTERVIEW ANALYSIS & SYNTHESIS

Interview transcripts were analyzed using Claude AI Skill that synthesized themes, quotes, pain points, and recommendations. These outputs fed directly into the PRD, ensuring product decisions were grounded in real user insights rather than assumptions.

KEY INSIGHTS

  1. Forward and backward navigation to give users more control while watching videos

  2. Short‑form news helps them keep up, but doesn’t help them understand

  3. News feels more trustworthy when it sounds human and conversational

FROM PRODUCT DEFINITION TO PROTOTYPE

PRD specifications and Figma design system constraints were provided to Claude Design as structured inputs, enabling AI‑assisted layout synthesis aligned with implementation requirements. The resulting UI/UX prototype was exported and handed off to Claude Code for further technical iteration.

A friend that tells you all you need to know in a way you want to hear it...

Combining vertical short‑form storytelling, a conversational AI journalist companion, and personalized narration grounded in trusted journalism.

Interested in more details ?

If you want to explore more about this and similar conversational experiences, feel free to reach out. I'm happy to share more insights privately. Connect with me on LinkedIn to discuss or stay updated on future projects.