Skip to Main Content

Key facts

There is currently no simple metric showing how productive people can be when supported by gen AI’s tools. Much depends on the task and the person. Our own initial experiments have shown that gen AI tools can reduce time spent on refactoring code by 20 to 30 percent and on generating code by 35 to 45 percent, but speed gains vary by task complexity and the developer’s experience.⁹ These tools perform best for relatively repetitive tasks and in providing a starting set of code that developers can work with and improve. While recognizing that true productivity is hard to measure, we’re seeing gains of 50 to 75 percent in productivity for developers with sufficient training (though we expect significant variance as the technology matures and new tools come online).

Gen AI tools can reduce time spent on generating code by

35–45%.

Implications

How your organization works will need to change. That can seem overwhelming. Start by identifying a business area and think through what changes are possible in how its work gets done. Pay particular attention to understanding which pivotal roles could benefit most from copilots. Nurture a strong learning culture and structured skill building that incorporate both formal programs, such as reviewing documentation, learning tools and code, and going to conferences, and informal programs, such as communities sharing information, making it easy to practice with new tools, and LLMs. 

⁹

“Unleashing developer productivity with generative AI,” McKinsey, June 27, 2023.

Powered by Ceros

⁹

“Unleashing developer productivity with generative AI,” McKinsey, June 27, 2023.

How your organization works will need to change. That can seem overwhelming. Start by identifying a business area and think through what changes are possible in how its work gets done. Pay particular attention to understanding which pivotal roles could benefit most from copilots. Nurture a strong learning culture and structured skill building that incorporate both formal programs, such as reviewing documentation, learning tools and code, and going to conferences, and informal programs, such as communities sharing information, making it easy to practice with new tools, and LLMs. 

Implications

35–45%.

Gen AI tools can reduce time spent on generating code by

There is currently no simple metric showing how productive people can be when supported by gen AI’s tools. Much depends on the task and the person. Our own initial experiments have shown that gen AI tools can reduce time spent on refactoring code by 20 to 30 percent and on generating code by 35 to 45 percent, but speed gains vary by task complexity and the developer’s experience.⁹ These tools perform best for relatively repetitive tasks and in providing a starting set of code that developers can work with and improve. While recognizing that true productivity is hard to measure, we’re seeing gains of 50 to 75 percent in productivity for developers with sufficient training (though we expect significant variance as the technology matures and new tools come online).

Key facts