Skip to content
All work
2026

AI Publishing System

Labelbox · Design & engineering

GTM velocity without brand drift

Overview

Marketing needed interactive blog widgets and full pages that still looked like Labelbox — but every custom page was weeks of eng. I owned the product from idea to implementation: an AI workflow constrained to approved design-system components that turns screenshots and briefs into CMS-ready experiences, so GTM can move at campaign speed without forking the brand.

Context

Interactive marketing pages were stuck behind custom eng work. Campaigns needed design-system fidelity at something closer to content speed.

How it evolved

Framed the product as a publishing workflow constrained to the design system — screenshots and briefs in, CMS-ready experiences out — and owned it from idea through implementation.

Where it is now

Weeks of production compressed to roughly 1–2 days. Case study covers the system depth.

What it delivered

1

Owned the product from problem framing through workflow design and implementation

2

Cut production from several weeks to roughly 1–2 days by encoding design-system rules into the publishing path

3

Deep dive on orchestration, review, and Contentful handoff in the AI publishing case study

How it ships

  • Gemini + Cloud Run for generation; Contentful as the publish target so marketing stays in their CMS, not in a custom eng pipeline

  • Constrained the model to approved components — AI accelerates shipping without inventing off-brand UI

  • Full case study: AI-Powered Publishing System

Surfaces

/case-studies/ai-publishing-system

Stack

Next.jsTypeScriptGeminiCloud RunGoogle CloudContentful

Team

Design & engineering ownership; idea through implementation