According to a new survey of U.S.-based developers by GitHub, 92% of respondents say they now use A.I. coding tools at work and for outside projects. The findings underscore how rapidly new generative A.I. tools are being adopted within businesses, albeit among a highly tech-savvy group of workers.

GitHub, which is owned by Microsoft, makes one of the most popular such tools. Known as GitHub Copilot, the “pair programmer” can explain code, make suggestions, and fix bugs with a recent integration of OpenAI’s GPT-4 model. Competition among A.I. coding tools is heating up, with Google announcing its competitor to GitHub Copilot at its I/O developer conference in May. And a number of similar A.I. coding tools are also available, including Amazon’s CodeWhisperer, Tabnine, and online platform Replit.

GitHub partnered with Wakefield Research to survey 500 U.S.-based developers at enterprise companies. The survey takers were mostly men in their thirties and forties at organizations with more than 1,000 employees, reports Gergely Orosz, author of the Pragmatic Engineer Newsletter. 

Many of the developers surveyed in the report said the tools helped them get into a better flow. “They feel more fulfilled because of the fact that they’re able to focus in on more meaningful work; on the reasons that they became developers in the first place,” said Christopher Harrison, senior enterprise advocate at GitHub.

GitHub’s chief product officer Inbal Shani described the benefits of today’s crop of A.I. coding tools as a major improvement from the start of her career decades ago, when she experimented with A.I. in the lab. Back then, it took her team’s models five days to process larger data sets. 

“I yearned for tools that would make me more efficient and shorten my time to production,” Shani wrote in an introduction to the survey.

Still, the spread of A.I. tools throughout organizations is raising concerns among some managers.

Google parent company Alphabet recently warned its employees not to make “direct use” of computer code with Google’s Bard A.I. chatbot, because the bot can make unwanted code suggestions, according to Reuters.

And while developers may feel a confidence boost from their assistants, A.I. tools haven’t always bested developers who rely on their own abilities. 

In a November 2022 paper, computer scientists at Stanford University looked at how users interact with an A.I. code assistant to solve a variety of security-related tasks. Those who had access to an A.I. assistant wrote significantly less secure code than those without access, yet were more likely to think they wrote secure code compared to those who didn’t use an A.I. assistant.  

Still, the researchers were hopeful that as improvements are made, the tools could result in higher productivity and bring other advantages like lowering barriers to entry and opening up accessibility in software development. GitHub recommended that companies establish standards for using A.I. tools so that they’re used ethically and effectively.