ANZ Bank: GitHub Copilot Boosts Developer Speed by 42%
Developer Productivity With and Without GitHub Copilot
42.36 percent faster
Headline outcome
Challenge
Establishing meaningful productivity metrics beyond simple coding time; Controlling for developer skill level differences in productivity measurement; Measuring long-term adoption versus short-term experiment results; Assessing code quality alongside speed improvements
Solution
A six-week longitudinal study at ANZ Bank found GitHub Copilot users completed tasks 42% faster overall, with beginners gaining a 52% speed improvement, demonstrating consistent AI productivity gains across all experience levels.
Outcomes
Overall task completion speed improvement: 42.36 percent faster (2023); Beginner developer productivity improvement: 52.27 percent faster (2023); Intermediate developer productivity improvement: 41.6 percent faster (2023); Advanced developer productivity improvement: 40.48 percent faster (2023)
Expert Insights
AI coding assistants show measurable productivity gains across all skill levels; Less experienced developers (beginners) see the greatest relative productivity improvements; Controlled experiments in real workplace settings provide more relevant results than lab studies; Task completion time is a reliable proxy for developer productivity in controlled studies; Productivity gains from AI tools vary by developer experience level but remain consistently positive
Discussion
Sign in to share your thoughts on this case.
No comments yet. Be the first to contribute.