Skip to content

SYSTEM Cited by 1 source

Zalando Content Creation Tool

What it is

Content Creation Tool is Zalando's internal UI used by photographers and copywriters to onboard new articles to the Zalando shop — uploading product images, enriching the article's attributes (colour, fit, neckline, assortment type, etc.), and performing the four-eyes-principle QA review before publication (Source: sources/2024-09-17-zalando-content-creation-copilot-ai-assisted-product-onboarding).

The tool is the human-facing endpoint of the content- onboarding pipeline. Images uploaded here flow into the Content Creation Copilot's Prompt Generator; attribute suggestions produced by the copilot flow back here as pre-selected values marked with a purple dot (concepts/ai-provenance-ui-indicator).

Role in the copilot workflow

  1. Photographer uploads product images; image URLs are sent to the Prompt Generator.
  2. Prompt Generator materialises an LLM prompt (using Article Masterdata for attribute schema), calls GPT-4o, runs the code↔English translation both ways, and returns structured suggestions.
  3. The Content Creation Tool auto-selects the suggested attribute values and marks them with a purple dot. The copywriter reviews (QA mode), un-selecting or adjusting as needed, then publishes.

This design — pre-select plus disclosure — is captured as patterns/pre-select-ai-suggestions-with-visual-disclosure. It explicitly shifts the human's workload from enrichment-then-QA to QA-only, which the post estimates removes ~25% of the content-production timeline.

Design properties

  • Four-eyes principle preserved. The tool has always required a human to approve every article before publication; the copilot does not change that contract. What changes is what the human starts from: an empty form vs. a pre-filled form with AI provenance indicators.
  • Uniform AI-provenance marking. The purple dot is applied identically to every AI-suggested value, regardless of which backend (systems/gpt-4, GPT-4o, future brand-data sources, partner integrations) produced it. See patterns/model-agnostic-suggestion-aggregator.
  • Final decision stays human. From the post: "the user interface clearly shows what suggestions come from which source, leaving the final decision in the hands of the human."

Seen in

Last updated · 501 distilled / 1,218 read