Producing on-brand B2B content for a global trade technology company using an iterative Claude workflow — matching a specialised industry voice from reference material alone.

Project Details

The Brief

Write a blog post that sounds like it came from inside the company

The client publishes authoritative long-form content on customs modernisation, government trade technology, and global trade policy — written for a sophisticated audience of government officials and trade professionals. The task was to produce a publish-ready blog post on their flagship AI-powered customs management system, matching the brand's existing tone and technical depth — without access to a brand brief or style guide.

The Challenge

Voice matching in a specialised industry

The client's content assumes the reader understands trade and government infrastructure at a professional level. Generic AI output — vague, overly structured, and full of filler — would immediately read as out of place. The challenge was to get Claude to produce content that felt genuinely informed, not just formatted correctly.

The Workflow

Reference-first, iterative prompting

1. Set up a dedicated Claude Project and uploaded existing company blog posts as reference material, giving Claude the context it needed to analyse tone, sentence structure, and content depth before generating anything.
2. Prompted Claude to internalise the brand voice from the reference posts first — not to summarise them, but to write in the same register.
3. Ran multiple drafting iterations, reviewing each against the reference material manually and refining the prompt based on where the voice drifted or where claims needed stronger grounding.
4. Applied editorial judgment at each stage — adjusting structure, adding real data points, and tightening language — before submitting the final draft to the marketing team for approval.

Tools used: Claude (Sonnet) · Claude Projects · Iterative prompting · Reference-based voice matching · Human editorial review