Research · Project
Digital Workers at the AI Edge
Online freelancers are among the first workers to both use and compete with AI. This new panel study follows how they adopt AI-enabled systems, and how both the workers and the systems change over time.
For more than forty years, work has been moving away from long-term employment toward contingent, project-based arrangements — and, increasingly, toward work mediated by digital systems. AI-enabled systems are the newest and fastest-moving of those systems. Online freelancers are the ideal place to study them, because they meet these systems on both sides at once: they adopt tools like ChatGPT, Copilot, and Claude to market themselves and do their work, and they compete directly with the same class of systems, since they are hired for exactly the discrete, knowledge-based tasks those systems are designed to perform. Full-time employees, whose jobs bundle many tasks and whose employers restrict which tools they may use, are more insulated — and harder to study. Freelancers are neither.
What we ask
The study pursues three questions: In what ways are workers leveraging AI-enabled systems? In what ways are they competing with them? And what are the trajectories of change in how workers use these systems over time? Answering the third is why the study is a panel: only by returning to the same workers year after year can we watch adaptation as it happens, rather than inferring it from a single snapshot.
How we work
The study continues and extends a longitudinal panel we have run since 2019. It pairs annual surveys and interviews with two panels — online freelancers and the employers who hire them — and adds a third strand that tracks the AI-enabled systems themselves as they evolve, alongside extensive secondary data on AI and employment. Each participant becomes a structured, de-identified case study; comparative analysis across cases and fuzzy-set qualitative comparative analysis (fsQCA) turn those cases into findings about which workers adapt, how, and to what effect.
Why it matters
Early evidence already unsettles the familiar debate. Workers are neither simply replaced nor simply helped: many adopt AI to gain an advantage, while clients — assuming the work is now AI-assisted — push to pay less, an attempt to capture the value of workers' own use of the technology. Because contingent, digitally mediated work is where workplace change tends to surface first, what these workers are living through now is a preview of what AI will mean for knowledge work more broadly.