At frontier artificial intelligence (AI) labs, the era of teaching AI models by scraping the internet and relying on the labor of low-paid workers to do simple annotation tasks (such as labeling by putting bounding boxes around motorcycles) is largely over. Today, these labs are purchasing complex demonstrations of white-collar work (particularly in the sciences, law, medicine, and academic research), in the hope of encoding expert knowledge and ability into the next generation of agentic models.
Demand for expert data has led to the emergence of a new gig economy for experts, who are hired as hourly contract workers by a few large data annotation firms that sell this data to the AI labs. While some experts may benefit from this arrangement, what will happen to expert careers, now and in the future, if expertise is turned into a "cheap" commodity – knowledge or ability that AI can replicate at low cost?
A Rutgers study examining these issues and the broader questions surrounding expertise, professional work, artificial intelligence, and AI-driven economic change was conducted by SC&I Assistant Professor of Library and Information Science Robert Wolfe, and co-author Aayushi Dangol, and co-author Aayushi Dangol, a researcher at the education research organization foundry10. The paper received the Best Paper Award from CHIWORK 2026, the Symposium on Human-Computer Interaction for Work.
Wolfe and Dangol explore what the expert data gig economy means for the future of knowledge work, including questions about who owns expertise, how professional knowledge is being captured and commercialized, and the roles played by highly skilled workers in training the next generation of AI systems.
"Our question in this study was how the firms collecting data from experts view the future of expertise and expert labor, and what this means for the millions of people who have invested in building skills for the information economy," Wolfe said. "So we set out to study how the firms and their CEOs described AI expertise, human expertise, and institutional expertise, and to begin a public conversation around what the future could or should look like for expert workers."
Wolfe and Dangol examined public communication posted to the X (formerly Twitter) feeds of five data annotation firms and their CEOs and then studied 29 podcasts on which those CEOs appeared between June and December of 2025. They analyzed the transcripts of those podcasts according to how the CEOs characterized AI expertise, human expertise, and institutional expertise and then produced three themes to describe each. Drawing on those themes, they posed a series of provocative questions intended for researchers, policymakers, industry leaders, and members of the public to productively inform future action.
The paper’s findings have broad implications for New Jersey residents and beyond, Wolfe said. "For generations, building expertise in some area has come with the promise of a comfortable middle-class life, one available not just to the best resourced people in our society but to anyone willing to work hard enough. Public institutions like Rutgers have served as remarkable engines of social mobility for precisely these reasons."
Dangol added, "Whatever the potential benefits of generative and agentic AI, the vision for the future of expert work put forth by the data annotation firms looks precarious, and more similar to the experiences of lower-wage workers in other tech-mediated gig economies. Tens of billions of dollars are being invested in this future, largely without the input of the public."
Looking ahead, Wolfe said, "We see a need to take seriously the question of what ‘cheap’ expertise would mean for society; even if the frontier labs manage to produce low-cost approximations of expert labor, what would our country or our state need to do to prepare for a world in which expertise is divorced from the expert? And if this is not a desirable future, then we need to begin the discussion of how to take steps toward ensuring that the public, including the people who have invested in building human expertise, benefit from the AI of the future."
