The debate over the effectiveness of AI as a coding tool may be missing the point, according to some experts. Inbal Shani, chief product officer for GitHub, believes that the real value of AI in coding lies in changing the nature of a programmer’s job. Shani argues that the introduction of natural language as an abstraction layer has opened up possibilities beyond code completion.
GitHub introduced its own code assistance tool, GitHub Copilot, in June 2021. Shani describes this year as “transformational” for AI in programming, with over a million paying customers and 37,000 organizations using Copilot. Accenture, one of the prominent users of Copilot, has seen a significant reduction in boilerplate code and an 88.5% retention rate of code generated by Copilot.
The use of Copilot has also led to a 15% increase in the number of pull requests completed on time at Accenture, indicating improved productivity. Developers are more willing to go through the build process, which builds trust and allows for more code to be deployed into production.
Shani emphasizes the qualitative benefits of these changes. By increasing the build rate, developers have more time to focus on architecture and other important aspects of software development. Shani highlights that developers have less than two hours a day to write code on average, as they are often engaged in other tasks such as testing, meetings, and writing pull requests. Automating some of these tasks frees up developers’ time and allows them to invest in other areas.
However, Shani acknowledges that the productivity increase has not been thoroughly quantified yet. The impact of tools like Copilot on developer happiness and recognition is also an important aspect to measure. Shani believes that measuring lines of code written is not necessarily an indicator of success or the quality of a programmer.
One of the significant advantages of AI in coding is the reduction in the need to switch between different tools. Copilot brings all the necessary information and search capabilities into the integrated development environment, eliminating the need for developers to switch between different tools and copy-paste code. This seamless integration leads to happier developers who can focus on their work without constant context-switching.
Copilot is also being used in various other areas of the programming team, such as coding interviews and onboarding new programmers. Shopify, for example, uses Copilot to assess new hires and educate new coders.
While Copilot and similar tools may not always produce the desired results, the potential for AI to transform the coding process and improve productivity is evident. However, more research and data are needed to fully understand the impact of these tools and define metrics for measuring success and developer happiness.