Asana cleared a stalled test-framework migration in two weeks with AI

Asana finished migrating its test suite off the Enzyme testing framework in two weeks, using AI coding agents to handle the bulk of the rewrite. Engineers there had pushed the migration off for years: rewriting a large body of test cases by hand was exactly the kind of mechanical, low-glamour work that keeps losing to feature work on a roadmap. Airbnb and Uber report similar outcomes on their own framework migrations.

The pattern matters more than any one company's timeline. Framework and dependency migrations are usually well-specified (the old API, the new API, a big pile of call sites) but tedious enough that teams delay them until a security deadline or an unsupported-version notice forces the issue. That's close to the best-case task shape for a current coding agent: pattern-based, reviewable in small diffs, with existing tests to catch behavior changes.

For a small team sitting on a similar deferred migration, the practical takeaway isn't "hand it to an agent and walk away." Test rewrites can silently narrow or drop assertions instead of translating them, and an agent under time pressure will happily produce a passing suite that verifies less than the one it replaced. Reviewing the diff test-by-test, not just watching the suite go green, is still the engineer's job.

None of the reports named a specific tool, model, or how much of the two-week window was agent time versus human review, so treat the "two weeks" figure as directional rather than a benchmark to plan a sprint around.