A current study by Uplevel shows that developer productivity does not increase noticeably through the use of AI coding assistants.

Over a three-month period, the output of 800 developers was compared with and without using Github’s Copilot coding assistant. The team was expected to write more code and review and improve less. Surprisingly, Uplevel did not see any significant improvements, according to Techspot. In fact, thanks to Copilot, there were 41 percent more errors in the code.

After all, programming and AI don’t fit together perfectly

AI coding assistants generally have little success because the workload does not decrease as hoped. The wizards often produce unsafe or difficult-to-understand code, forcing developers to spend more time debugging than writing the code themselves. Nevertheless, AI is often used to review or improve code. This is likely due to the rapidly growing range of open source coding assistants and algorithms that continually learn.

Companies like Innovative Solutions are happy with the results, with productivity tripling thanks to tools like Claude Dev and Copilot. Obviously, coding with AI is still in its infancy Success rate is not the same everywhere, but at the current rate of development this can change very quickly.