A team at Atlassian built a production-ready frontend feature (Public App Updates) in 7 weeks using RovoDev, compared to a 6-month manual estimate — over 3.5× faster. The post shares the practical workflow: building a project knowledge wiki to reduce redundant AI research, using memory files for continuous improvement, generating well-defined Jira tickets via AI, and layering code quality guardrails (TypeScript, unit tests, integration tests, VR tests, E2E). Key tips include pointing AI to concrete examples, using skills to keep context small, organizing models into Advisor/Executor roles to manage cost, and giving AI 'eyes' via visual regression tests and Playwright. The author argues experienced engineers are more valuable than ever as 'AI engineers' who can guide AI through complex systems and ensure production-quality output.

11m read timeFrom atlassian.com
Post cover image
Table of contents
How it startedKnowledge graphJira tickets as tasks, memory, and audit logsCode quality guardrailsEngineering knowledge is more valuable than everPractical tipsImpact?Areas to improve in your team

Sort: