AI-powered copy editing system

Launching a new product at an international organic food brand means generating a ton of copy across multiple languages and channels while staying compliant with EU and national level food labeling and green claims regulations.

Before a single word is written, the marketer has to gather all relevant data points about the new product- benefits, interesting ingredients, cultural backgrounds - from various stakeholders in the company, especially the product development team.

AI
Claude

That copy then needs to be prepared to work across the official product pages, social media, product sheets, and exposition catalogs, each with its own length and style requirements.

The old workflow was complicated: the marketer would write a draft, hand it over to the QM department for review, and then have to manually resize the text to fit several dozen different length and style requirements. Repeat for 2-6 languages. It was slow, error-prone, and didn't scale well.

What the system does

Introduced an AI-based system for systematic product briefing and copy generation, built around four capabilities:

Structured data collection - helps gather all data points that need to go into creating the product copy, with priorities, before writing begins
Completeness check - reviews human-written copy to verify that all required data points are actually covered

Preliminary compliance check - flags potential issues with food regulations, labeling requirements, and green claims before the copy reaches the QM department
Text resizing - cuts the approved copy into the many size and format variants needed across channels

This puts powerful tools in the hands of the marketer to speed up the most time-intensive parts of their workflow - but leaves the text creation itself in human hands.

Built on the command line

The workflow currently runs in the terminal using Claude Code, but is fundamentally provider agnostic - it can run with any AI provider. Product information briefings, generated copy, quality check results, and compliance flags are all plain Markdown files in the local file system.

Nothing is locked into an AI provider's interface or proprietary format. The process stays auditable , portable, and easy to integrate into existing workflows.