OpenAI’s Codex Helps Asana Cut 5-Year Migration to 2 Weeks, Slashes Costs 500x

OpenAI’s Codex helped work management platform Asana migrate from the outdated Enzyme testing framework to React Testing Library (RTL) in about two weeks—something the company had estimated would take at least five years. 

In a blog post, OpenAI said that the project cost about $12,000 in model and infrastructure usage, compared to roughly $6 million that Asana had previously estimated.

Enzyme had fallen out of active maintenance and had been blockingAsana’s efforts to modernise its frontend stack. 

Rather than manually migrating the codebase over several years, Asana simultaneously used up to four Codex coding agents, with each agent working from a separate copy of the codebase.

The project started with a five-sentence prompt. 

An engineer checked progress twice a day and reviewed every proposed change, retaining human oversight over the migration. Asana said simpler instructions worked better than a more elaborate agent setup.

After 1.5 weeks of engineering effort spread across two calendar weeks, Enzyme had been fully removed, and the migration to RTL was complete. 

The migration gives Asana a new way to approach other large engineering projects that had previously been considered too time-consuming to justify. 

The company now plans to test coding agents on additional migrations, rewrites, and performance problems.

“Not every years-long project will collapse into weeks,” Asana CTO Amritansh Raghav said in a blog post. “But agents can give engineers more room for craft—and make once-impossible work worth attempting.”

Codex’s enterprise adoption has expanded rapidly beyond individual developers. 

Recently, OpenAI said weekly active enterprise users have grown 5x in engineering since February, while usage increased 108x in legal, 41x in sales, 41x in recruiting, and 26x in marketing.