Cursor at Grab: Adoption and impact
Article Summary
Grab achieved 98% monthly active adoption of Cursor across their engineering org in just months. That's nearly 30 points higher than industry benchmarks for high-performing teams.
Grab's engineering team shares how they scaled AI coding assistants across 800+ cities in Southeast Asia. They took a multi-tool approach rather than betting on a single solution, and the results speak for themselves.
Key Takeaways
- 50% suggestion acceptance rate, beating the 30% industry average significantly
- Over one-third of merge requests now use Cursor in some capacity
- Non-engineers now ship production code: designers merged hundreds of PRs independently
- Custom monorepo indexing and Grab-specific rules made suggestions actually relevant
- Tasks that took full days now complete in hours
Grab turned an AI coding tool from experiment to daily workflow for 98% of engineers while expanding coding capabilities to non-technical teams across FP&A, ops, and design.
About This Article
Grab needed to figure out if Cursor actually made engineers more productive, or if the numbers just reflected more usage. They had to separate the tool's real impact from other factors that could affect output.
Grab used fixed-effects regression analysis to measure how Cursor affected productivity while accounting for confounding factors. This let them establish a clear relationship between how much people used Cursor and how much they produced.
The early results showed that productivity gains scale with Cursor usage intensity. The effects held up under statistical scrutiny across Grab's engineering teams in multiple countries.