Asana replaced a legacy testing infrastructure in 14 days using OpenAI Codex, finishing a project scoped at five years of engineering work for roughly $12,000 total.
The number that matters is not the speed, it is the cost-to-scope ratio. Five years of engineering labor compressed into two weeks at $12K exposes how dramatically AI coding agents can reprice large-scale refactoring work. The original article details the specific Codex workflows, prompt strategies, and human oversight checkpoints Asana used, and those mechanics are worth studying before drawing conclusions about what is replicable.
This is an early, controlled case study from a company with strong engineering culture and a well-defined legacy target. The real question the full piece raises but does not fully answer: what breaks when teams without that foundation try the same approach.
[READ ORIGINAL →]