Asana Migrates Five-Year Legacy Project in Two Weeks Using OpenAI Codex
A $6 million engineering estimate was reduced to a $12,000 task, signaling a shift in the economics of software maintenance.
Asana has migrated its frontend testing system from the outdated Enzyme library to a modern stack using OpenAI Codex. The project, previously estimated to require five years of manual engineering effort, was completed in approximately two calendar weeks.
To execute the migration, Asana deployed up to four AI coding agents working in parallel on separate copies of the codebase. A human engineer oversaw the automated process to ensure quality and accuracy. The total expenditure for the model and necessary infrastructure was approximately $12,000. In contrast, the company's previous manual staffing plan for the same migration was estimated to cost roughly $6 million and take at least five years to complete.
The Burden of Technical Debt
Enzyme is a JavaScript testing utility for React that fell out of active maintenance. For a company with a large-scale codebase like Asana, the lack of updates for such a foundational tool creates a significant blocker for modernizing the broader frontend stack. This situation often results in "technical debt," where the sheer volume of manual work required to replace a legacy system becomes prohibitively expensive, forcing companies to avoid necessary upgrades.
A Shift in Engineering Economics
This case study demonstrates a massive shift in the economics of software maintenance. By reducing a multi-million dollar, multi-year project to a two-week task, Asana has shown that legacy migrations previously deemed "impossible" or too costly can now be executed rapidly. This allows engineering teams to pivot their focus away from tedious maintenance and toward the "craft" of developing new features.
Asana CTO Amritansh Raghav noted that while not every long-term project will collapse into a matter of weeks, agents provide engineers with more room for craft and make previously impossible work worth attempting.
The Future of Codebase Rewrites
The success of this migration changes the strategic calculation for when and how companies perform large-scale codebase rewrites. Rather than managing decaying legacy systems for years, firms may now opt for AI-driven migrations to maintain technical agility. The industry will now be watching to see if this model of parallel AI agents can be scaled to more complex architectural shifts beyond testing frameworks.