A product marketer at Atlassian shares how cross-functional teams built and validated 'golden prompts' for Rovo, Atlassian's AI content creation tool in Confluence. Golden prompts are carefully crafted, rigorously tested prompts designed for high-stakes, repeatable use cases across personas like Marketing, Sales, Product, and Engineering. The team tested outputs across multiple AI tools (ChatGPT, Claude, Gemini, Perplexity, Rovo), documented results in Confluence, and used Confluence Whiteboards to map end-to-end demo scenarios. The post includes concrete prompt examples per persona and a five-step evaluation process for refining prompts until outputs meet a live-demo quality bar.
Table of contents
What is a “golden prompt” — and when should you use one?Why we started with golden prompts (with examples)How we defined prompt “quality” — and why it matteredDesigning real‑world scenarios with Confluence whiteboardsWhy this work went beyond “a good demo”Cross‑functional partnership: our unfair advantageSort: