Asana cleared 5 years of engineering work in 2 weeks with Codex

Asana, a work management platform, replaced its outdated testing system Enzyme using OpenAI Codex in approximately two weeks. The project, which involved removing Enzyme to modernize the frontend stack, was completed at a cost of about $12,000. Asana’s engineers used Codex, powered by frontier models, to automate large codebase changes, with up to four coding agents working in parallel. An engineer reviewed progress twice daily and approved each proposed change throughout the process.

The migration was driven by Enzyme’s lack of active maintenance, which had become a barrier to upgrading Asana’s frontend. Using a five-sentence prompt, the team leveraged Codex to decommission the system efficiently. Simpler instructions proved more effective than a more complex setup. After 1.5 weeks of engineering effort spread over two calendar weeks, Enzyme was fully removed from the codebase.

Model and infrastructure costs for the project totaled around $12,000, a significant reduction compared to the previous staffing plan’s estimated $6 million expense. The previous plan had projected a timeline of at least five years to complete the migration. Asana’s leadership noted that the project’s success shifted perceptions of which long-running software initiatives are now feasible.

With Enzyme removed, Asana plans to apply Codex to other migrations, rewrites, and performance challenges previously considered impractical. The team aims to free engineers to focus on higher-level work while maintaining human oversight of automated processes. This approach reflects Asana’s broader strategy of integrating AI agents into its engineering workflows.