Artificial intelligence is expected to usher in productivity gains on par with the Industrial Revolution, but economists don’t always agree what kind of impact it will have on the labor market. While most believe AI will eliminate jobs, the exact scope and speed of this displacement, which roles are most likely to be automated and whether or not the technology will eventually create more jobs than it eliminates are all points of contention.
Why No One Can Agree on AI’s Impact on Jobs
Studies about AI’s impact on jobs often reach different conclusions because they all analyze different data sets, industries and labor market outcomes, such as hiring, layoffs or productivity. Researchers also use different methods to measure AI exposure, and businesses are adopting the technology at different speeds, making comparisons difficult.
With every new study released, the picture becomes less clear. And when a consistent theme emerges from the data, new research pops up to poke a hole in those narratives. The inconsistencies are not only confusing, but they also raise bigger questions about how we collect and analyze AI-related jobs data, which will ultimately affect how we prepare workers and society for an economy defined by artificial intelligence.
We Want to Know
A Bright Spot in Entry-Level Hiring
If there have been any clear conclusions from the dozens of studies on AI-related job loss published over the last couple of years, it’s that it will have a disproportionate impact on recent college graduates applying for entry-level roles. Since the launch of ChatGPT in 2022, employment among early-career workers in AI-exposed occupations has dropped 16 percent, according to a Stanford University study. And as of March 2026, recent college graduates are now more likely to be unemployed than other Americans. Over the past year, Goldman Sachs economists estimate AI has reduced U.S. monthly payroll growth by roughly 16,000 jobs, hitting younger, less-experienced workers the hardest.
But that narrative is now being challenged by expense management company Ramp, which linked data about its customers’ AI spending with Revelio Labs’ workforce records to reveal that the companies investing heavily in artificial intelligence actually hired more employees. According to the June 2026 study, high-intensity AI adopters grew their teams by 10.2 percent within two years of adoption. All business functions, from customer service to engineering, saw headcount growth, and the number of entry-level workers increased by 12 percent. Low-intensity AI adopters, meanwhile, did not see a meaningful change in employment. Researchers said this study counters the prediction that AI will eliminate jobs in general, and entry-level jobs in particular.
Mixed Signals in Software Engineering
Perhaps the second-biggest takeaway from AI-related jobs surveys has been the softening demand for early-career software engineers in particular. Stanford’s 2026 AI Index found that employment for software developers between the ages of 22 and 25 has fallen nearly 20 percent since 2024. And according to 2024 data from the Federal Reserve Bank of New York, first-time job seekers with a software engineering or computer science degree are more likely to be unemployed than their peers with art history or liberal arts degrees.
While many predicted that the rise of AI coding tools and vibe coding would reduce the demand for software engineers, the profession has proven more resilient than previously thought, according to a study by venture capital firm SignalFire. While hiring at major tech companies has dropped 25 percent since 2019, engineering hiring has only decreased 11 percent. The study also found that the proportion of new hires at major tech companies has increased from 46 percent to 55 percent.
However, entry-level hiring was less promising, dropping roughly 65 percent at major tech companies and about 76 percent at early-stage startups. AI may be enabling these new grads to start companies of their own, though, with the report noting that the top computer science graduates of 2025 were twice as likely to call themselves a founder compared to the class of 2022.
Conflicting Data Complicates AI’s Impact
The numerous conflicting studies highlight the difficulties of measuring AI’s impact across numerous business contexts in real-time.
First, these studies are based on private data sources focused on specific segments of the workforce. Ramp’s subjects, for example, are larger tech companies that are already growing fast, often with the help of venture capital. Those companies may be experiencing growth, including AI productivity gains, that are not representative of the larger business world. Still, private data sources, like Anthropic’s Economic Index, are incredibly useful in providing a real-world snapshot into how businesses are actually using AI.
Second, almost every study looking at AI-related employment trends attempts to measure the “AI exposure” of various occupations by tracking the technology’s theoretical capabilities to real-world jobs by relying on the government’s O*NET occupational database. These exposure scores can vary widely depending on the survey’s methodology or the model used to analyze the data. When researchers at Northwestern University and American University measured AI exposure using three different frontier models, they reached drastically different conclusions about how AI is reshaping those professions.
Lastly, the conflicting data sources could be indicative of the varying approaches businesses have taken with AI adoption. As evidenced by the Ramp-Revelio Labs study, some companies may be experimenting with a chatbot subscription, while other companies have integrated AI into their operations, found efficiencies and grown their business to the point of needing more workers. Self-reported efficiencies can also be unreliable. In fact, some experts believe that executives are disingenuously citing AI in layoff announcements because the real drivers — like tariffs or overhiring during the pandemic — may spook investors.
Better Government Data Is Needed
While researchers have tried their best to gather the most relevant data, the most comprehensive data sets — those collected by the government — have notable limitations. Legislators, AI labs and economists have all called for better government data to ensure they are getting more comprehensive information about AI-driven shifts in the workforce.
The Bureau of Labor Statistics’ monthly jobs report, for example, uses broad industry classifications rather than a dedicated tech category, grouping many roles under “information,” which also includes publishers, broadcasters and telecommunications. And the few government surveys that ask employers about AI do so infrequently and suffer from significant processing delays. To make matters worse, the Bureau of Labor Statistics has struggled with declining survey response rates, which can hinder data quality.
Getting accurate labor data has never been more important with the introduction of AI tools, so in June 2026, a bipartisan group of U.S. senators proposed a bill, the Artificial Intelligence Data Authorization and Transparency Act, that would help the Bureau of Labor Statistics and other agencies modernize its surveys and better analyze AI’s effect on the workforce. Other bills have also called for better AI-related workforce data, including one that would require government agencies, publicly traded companies and some private companies to share quarterly data about AI-related layoffs.
The federal government is reportedly taking steps to address the problem. As part of Trump’s AI Action plan, the Department of Labor is developing an “AI workforce hub” that would provide empirical evidence — including new sources of private sector data — that highlights AI’s impact on the labor market.
If AI has the effect that economists expect it to, we will need comprehensive real-time data that measures its influence across all types of workers, companies and industries. This data will be necessary to help people transition to new careers and provide early warning indicators to government officials. Without more timely, comprehensive data, AI could displace employees and shock the labor market before the government has time to develop an equitable safety net.
Frequently Asked Questions
How do researchers determine which jobs are most affected by AI?
Researchers often estimate AI exposure by comparing the capabilities of artificial intelligence systems with the tasks performed in different occupations. Many studies rely on public occupational databases such as O*NET, but exposure estimates can vary depending on the methodology and AI models used.
Why is better government data needed to measure AI’s impact on jobs?
Existing government labor data was not designed to track AI-related changes in the workforce. Current surveys may not provide timely information about AI adoption, and some government employment categories combine technology jobs with unrelated industries, making AI’s effects harder to isolate.
What could better AI-workforce data help policymakers understand?
More comprehensive AI-workforce data could help policymakers identify changes in employment trends, understand which workers and industries are most affected and develop programs to help workers transition into new careers.
