Stanford's Canaries Dashboard Shows AI Is Already Squeezing Early-Career Workers Out

Picture two recent graduates, both hired into tech roles the same year. One lands in a position where AI tools augment her judgment; the other steps into a role where AI replaces the tasks that used to justify his entry-level salary. Three years later, the divergence between them is not theoretical. It is showing up in payroll records. That gap is now the subject of one of the most carefully constructed empirical projects in workforce research.

What the Canaries Dashboard Actually Measures Stanford's Digital Economy Lab

built the Canaries Dashboard in collaboration with ADP Research, drawing on anonymized payroll data from firms that use ADP's payroll services, according to the Stanford Digital Economy Lab project page(opens in new tab). The dashboard's name comes directly from the underlying research: a 2025 paper by Erik Brynjolfsson, Bharat Chandar, and Ruyu Chen titled "Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence." The logic of the metaphor is precise. Certain workers, specifically those who are early in their careers and concentrated in occupations with high AI exposure, function as leading indicators for broader labor market shifts. Where they go, the rest of the economy may eventually follow. The dashboard updates these findings on a live basis, placing it alongside other labor market indicators in Stanford's expanding AI Economic Indicators series. The underlying paper provides the analytical spine. Brynjolfsson, Chandar, and Chen studied how employment changes vary depending on a worker's experience level and the degree to which their occupation is exposed to AI-driven automation. According to Bharat Chandar's October 2025 post on the Stanford Digital Economy Lab site, the research provided "some of the earliest large-scale evidence consistent with the hypothesis that the AI revolution is beginning to have a significant and disproportionate impact on entry-level workers in the American labor market." That framing matters. This is not a projection model or a thought experiment; it is a pattern found in payroll data that already exists.

The 13 Percent Signal and What It Means The finding that stops readers cold is a

13 percent relative employment decline among the most AI-exposed young workers, even after controlling for firm-level and time-based factors, according to a summary of the Stanford research. That figure is not an aggregate across all workers. It is specific to the intersection of two conditions: being early-career and being in a highly automatable role. Workers who are more experienced in the same occupations do not show the same pattern to the same degree, which is a crucial distinction. AI tools, at least in this period, appear to be substituting for the tasks that entry-level workers are hired to learn, not for the judgment that more experienced workers are paid to apply. The occupational distribution of this effect is also telling. According to a summary of the Stanford findings reported in Time, the steepest losses cluster in software development, customer support, and other roles with high AI exposure. These are precisely the roles that a generation of career-advice content pointed to as stable, skill-building entry points. The compression is hitting hardest where early-career workers were told to start. There is a second finding layered beneath the employment numbers: the mechanism matters enormously. Where AI replaces tasks outright, entry-level roles dry up. Where AI augments human work rather than replacing it, the effect is far less severe, according to the same Stanford summary. That distinction between replacement and augmentation is not a subtle academic nuance. It is the fork in the road that determines whether a given role grows or shrinks as AI adoption deepens.

The Difference Between a Wage Problem and an Exclusion

Problem Most public discussion of AI and work gravitates toward wages: will AI lower pay? The Stanford research reframes that question. According to the summary of Brynjolfsson, Chandar, and Chen's findings, the primary risk for the most exposed young workers is not lower wages. It is exclusion from employment altogether. Wage suppression and job loss are related problems, but they are not the same problem, and they do not call for the same response. A worker who gets hired at a lower wage still builds experience, still accumulates tenure, still develops the judgment that raises their value over time. A worker who never gets the entry-level position has no runway at all. This distinction should shape how learners evaluate their own positioning. The question is not simply whether a role pays well today. It is whether the role still exists in volume at the entry level, whether AI is replacing the foundational tasks within it, and whether the experience on offer actually builds the kind of judgment that is harder to automate. A role in a high-exposure occupation is not automatically a dead end, but the evidence now says the entry point to that role is narrowing.

What This Means

If You Are Deciding Where to Invest The Canaries Dashboard is part of Stanford's AI Economic Indicators series, which the lab describes as an effort to connect "policymakers, business executives, and individual workers to timely and reliable information on the economic impact of AI," according to the lab's indicators page. That last category, individual workers, is the one that matters most for anyone reading this. The dashboard is updated on a live basis, which means the 13 percent figure is a baseline to watch, not a final verdict. For learners assessing career paths right now, the actionable read is this: exposure level and experience level interact. A junior software developer in a firm deploying generative AI for code generation is in a different position than a junior data analyst whose work involves synthesis, stakeholder communication, and domain interpretation. The former role is closer to the tasks AI currently replaces; the latter is closer to the tasks AI currently supports. That difference is not captured in job titles, and it is barely captured in most job descriptions. The Canaries Dashboard and its underlying research give learners a framework to ask better questions about which entry points are genuinely building toward something, and which ones are quietly disappearing from the labor market before the job ads catch up. Monitor the dashboard as it updates monthly. Watch whether the 13 percent figure widens, narrows, or migrates into new occupational categories. The research is early, as Chandar has acknowledged, but it is the most honest signal currently available. That is worth more than any forecast.