Businesses today are racing to adopt generative AI into workflows to stay competitive. Research shows these tools help boost productivity, yet as companies look to reshape their workforces with AI, less is known about how it’s altering job roles in the information economy. When workers use GenAI, does it also change the way they work?
In the working paper Generative AI and the Nature of Work, researchers aimed to shift the GenAI conversation away from how much workers are doing and examine how AI changes what they’re doing.
When people used GenAI tools, researchers found, tasks shifted away from managerial work toward more core work, developers worked more independently, and they spent more time exploring projects that were unfamiliar to them.
Finding the right developers for the job
The research, conducted by Frank Nagle at the MIT Initiative on the Digital Economy and the Advising Chief Economist at The Linux Foundation, along with researchers from GitHub, Microsoft, the Linux Foundation, and UC Irvine, focused on a group of open source software developers using GitHub Copilot during its initial general access launch.
GitHub Copilot is a generative AI coding tool that functions as an assistant, teacher, and problem solver that suggests code snippets, assists with code testing, and helps developers learn new coding languages.
The researchers wanted to focus on workers who were decentralized and open source developers typically work in decentralized settings. These developers are also often overburdened, spending more time than they’d like on project management and less time than they’d like on actual coding.
The researchers wanted to know, if developers saved time on day-to-day tasks, how would they spend it? And if they got the answers to coding challenges from a chatbot, how would that affect the way they collaborate with others?
Examining Early Adopters of GitHub and Microsoft Tech
Between 2022 and 2024, the study examined the coding work of more than 187,000 software developers using GitHub Copilot, a GenAI tool for software development. Tasks included not only hands-on development tasks, such as writing code, fixing bugs, and conducting testing, but also managerial tasks, such as discussing bug fixes with other developers and reviewing requests for changes.
They found that the developers that got free early access to GitHub Copilot had a 12% increase in the percentage of tasks spent on core work and a 25% drop in the percentage of managerial tasks developers did outside of writing code.
When developers save time, the research showed they reinvested it into hands-on coding tasks.
At a recent IDE Lunch Seminar, Nagle said the shift in task allocation is rooted in economics.
“When it becomes cheaper to do core work, people are going to do more of it,” said Nagle. “Which is interesting because some of the literature on AI argues that we’re all going to have all this free time because AI is going to do our job. As it turns out, economics would predict the exact opposite. We’re all going to do more work because it’s cheaper and we can do more, at least in some contexts.”
Working independently, less collaboratively
In addition to task reallocation, GenAI also changed how developers approached tasks. The data collected on GitHub Copilot users showed more tasks were done independently, rather than collaborating with other developers. This indicates that when developers ran into problems or had questions, they turned to GitHub Copilot, instead of seeking advice and solutions from peers.
More independent work doesn’t always mean better results, said Nagle. Collaboration is an important part of coding–and arguably any role–especially for less experienced workers. Asking a colleague for help might take time from core tasks and reduce productivity, but shifting to a more independent work model could limit opportunities for mentorship and developing innovative ideas.
“As we think about the broader implications of the technology as a whole, this is one we’re going to have to manage as a society, this reduction in collaboration,” said Nagle. “Yes, it’s great that you can do more by yourself, but this may have more long-term implications than we realize.”
Using AI to advance careers
In today’s tough job market and growing worker anxiety about AI affecting head count, GenAI could offer worker benefits, such as helping them learn new skills and advance their careers.
Nagle noted that the shift toward core coding tasks was even more pronounced for lower-ability developers. When less experienced workers use GenAI, it bridges knowledge gaps and saves time workers would traditionally spend asking senior developers for help.
This is where business leaders should take note. AI offers additional value by upskilling entry-level workers, helping them develop into more senior roles faster, especially when junior employees are partnered with experienced mentors.
“This approach also presents a powerful strategic opportunity,” Nagle wrote in Fortune. “While your competitors are under-investing in their future, you can attract and hire the best of the next generation.”
Developers using GitHub Copilot also increased the percentage of tasks spent on exploration. Instead of diving deeper into the project they were working on, they invested that time on starting new projects or working in new programming languages. Having a novel project or exploring a new language can be fun and engaging–which in turn could mean developers are more satisfied in their jobs.
This kind of exploration could boost their careers, Nagle said, and a back-of-the-envelope calculation showed an increase in earning potential too. Writing code in certain languages earns higher salaries. Developers who use tools like GitHub Copilot for language learning could increase their labor market potential by $1,683 per year.
AI and restructuring organizations
More research is planned, Nagle said, to understand the time spent on tasks and how work outside of GitHub Copilot affects developers’ processes. Even when acknowledging the need for deeper context, this initial research offers new insights into the way AI is reshaping workplaces and processes.
If workers can use GenAI to do more core work and less coordination, operate more independently, and have more bandwidth for exploring new technologies, then it will likely level the hierarchical playing field within corporate structures. Workers can learn new skills with less formal training and mentoring. In turn, managers who have moved away from core work can dive back into these tasks, creating a more collegial work environment.
When companies recognize this deeper level of GenAI transformation, they can better align roles and prepare for the workplace of the future.