Virtually everyone agrees that artificial intelligence has the potential to reshape the economy in the coming decades. But no one is sure what effect the technology is having right now.
By some measures, AI is contributing to high graduate unemployment rates and may have already destroyed tens of thousands of jobs. Other sources suggest that companies are actually adding workers as a result of the technology.
AI can contribute to the US inflation problem or part of the solution to it. It may be responsible for a recent surge in productivity growth, or it may play almost no role—or the productivity boom itself may be a mirage.
Scientists can’t even agree on basic questions like how many companies are using AI or which workers are most vulnerable to the disruptions it could cause.
The conflicting signals partly reflect the challenge of detecting economic shifts in real time. Public statistics are inherently backward-looking and are better at measuring broad trends than developments in specific sectors or regions. New technologies that can lead to the emergence of new products, jobs or entire industries can be particularly difficult to measure.
What makes AI different is the speed of its diffusion through the economy. It has taken less than four years for generative artificial intelligence to go from a novelty mostly useful for writing limericks to a powerful tool adopted by the world’s largest companies. Economists have become convinced that the technology will have profound implications for workers and the economy, though they disagree about what those implications will be. Once the data is clear, they warn, it may be too late for policymakers to figure out how to respond.
“The stakes are super high,” said Nathan Goldschlag, director of research at the Economic Innovation Group, a think tank. “Getting the policy right is going to depend on getting the measurement right. You can’t get the policy right if you don’t know what’s going on.”
Mr. Goldschlag published a report on Thursday that documents the challenge of AI measurement and suggests steps to improve it. He and other experts argue that the government and the private sector should devote more resources to the problem.
At least they get a hearing in Washington. In June, a bipartisan group of senators introduced a bill that would expand data collection and require the federal government to produce an annual report on AI’s impact on the workforce.
“The government has to make some big decisions about artificial intelligence and about the economy, and if you do it in a vacuum, you’re going to make mistakes,” said Senator Mark Kelly, an Arizona Democrat and one of the bill’s sponsors. “This affects the lives of millions of Americans and millions of businesses. And you can’t do this smartly without reliable data.”
Mixed signals
Politicians are not flying blind. Since 2023, the Census Bureau has asked companies about their use of artificial intelligence in a survey every two weeks. It has also included questions about the technology in an annual business survey, though only intermittently.
Researchers have developed several measures of “AI exposure,” many of which use a public database of job descriptions to assess which occupations will be most affected. Economists can use these measures to find out whether the most exposed professions e.g. adding jobs more slowly or experiencing different wage increases.
The problem is that the sources often tell confusing or contradictory stories. Surveys arrive at completely different estimates of companies’ AI use based on how questions are asked. AI exposure metrics tell different stories about which jobs will be most affected. In one study, economists at Northwestern University and American University found that when they used different exposure measures, these could affect not only the magnitude of AI’s impact on jobs, but also its direction. AI hurt employment by some measures and helped by others.
“It’s like going to the doctor and getting three different diagnoses for the same condition,” said Michelle Yin, an economist at Northwestern University who was one of the study’s authors.
Part of the problem is that the best-known measures of the economy were developed for an era before personal computers and the Internet, let alone AI. The monthly jobs report from the Bureau of Labor Statistics, for example, provides breakdowns of job growth in manufacturing, retail and construction, but not in technology, which has adopted AI tools most aggressively. Instead, technology is spread across several categories, including information, a broad sector that also includes newspapers and film studios.
The jobs report provides even less information on occupations that may be vulnerable to displacement, such as software developers, accountants and customer service agents. The latest breakdown of detailed occupations is from May 2025, an eternity ago in the rapidly evolving world of AI
Still, economists say that for all its shortcomings, government data will be crucial to understanding AI’s impact over time. Researchers at the Yale Budget Lab, for example, have begun publishing a monthly analysis based on government data that tracks “job attrition,” how quickly the types of jobs that make up a given industry are changing. The measure is designed to be something of an early warning system for AI’s effects. As companies begin to adopt the technology, the researchers theorize, they will likely begin hiring for different roles, even if their overall headcount does not change immediately.
“It’s easy to collect case studies in retrospect,” said Martha Gimbel, executive director of the lab. “What makes this time different is that we’re actually trying to measure this and figure it out in real time.”
But those efforts may be hampered by a federal statistical system that has been plagued by declining response rates to government surveys. Shrinking budgets have made it difficult for statistical agencies to fill the gaps. Erika McEntarfer, who led the Bureau of Labor Statistics until President Trump fired her last year, said an additional $10 million a year in funding would allow the agency to expand the sample size of its monthly labor market survey so it could do a better job of capturing economic shifts.
“The data we currently use to understand the impact of AI on the labor market is at risk due to a lack of funding,” she said. “It would only take some very modest investment to prop them up.”
Private data
Many economists are not waiting for the government to catch up. Several research teams have released AI measures based on data from the private sector that are more detailed and more timely, albeit less comprehensive, than what is available from the government.
The Stanford University Digital Economy Lab last month released a dashboard of AI indicators based in part on data from ADP, the payroll processor. This data shows that entry-level jobs have fallen sharply in the most AI-exposed sectors since ChatGPT debuted in 2022. Erik Brynjolfsson, the lab’s director, called the trend a canary in the coal mine of AI-driven job losses.
“I think it’s comparable to the industrial revolution in terms of how it’s going to affect the labor market,” Mr Brynjolfsson said. “I wish the federal government would invest more in that. But in the meantime, there are some great private data sources that we’re gathering, and that’s what I think is helping to fill that gap.”
But the private data is as muddy as the government statistics. Research published this week by Ramp, a cost management firm, and Revelio Labs, a labor market data firm, found that the companies that used AI the most added jobs faster than those that had been slower to adopt the tools.
Ramp has access to data about which AI tools their customers buy and how much they spend on them. That makes it possible to distinguish heavy users from more cautious users — a crucial distinction because it takes time and investment for companies to figure out how to use the tools effectively, said Ara Kharazian, chief economist at Ramp.
“It’s hard to measure AI’s impact on a business because it requires sustained adoption,” he said. “It is clear in our work that a simple chat subscription does not promote productivity for a company.”
However, such data are not necessarily representative of the entire economy. ADP’s clients tend to be relatively large and well-established. Ramp’s customers tend to be tech-savvy. But if AI is to have the effects its biggest boosters are promising, it needs to be adopted by businesses of all shapes and sizes.
Work in progress
Researchers generally agree on one thing: AI’s effect on the broader economy has so far been limited.
This is not necessarily surprising. Mr. Brynjolfsson and other economists have found that technological innovations often follow a J-shaped pattern, with companies initially becoming less productive when experimenting with new tools, then experiencing rapid gains when they figure out how to exploit them.
The confusing economic evidence suggests that many companies are still on the downward leg of the J.
“The signals are mixed because the underlying economics are probably mixed because we’re still in a period of experimentation,” Mr. Goldschlag, the economist at the Economic Innovation Group. “The tools themselves are still becoming useful.”
If white-collar jobs really do start disappearing en masse, as some in Silicon Valley predict, it won’t be long before the losses show up in government data. But even then, it may not be obvious that AI is to blame.
The US economy has gone through a series of shocks in recent years that have nothing to do with artificial intelligence: the Covid-19 pandemic and its knock-on effects, including struggles to return to the office that continue to this day; inflation and the high interest rates adopted by the Federal Reserve to combat it; and drastic swings in government policy on immigration, trade and other areas. If a company has cut jobs since 2022, it is not easy to say whether it is the result of artificial intelligence, high interest rates or both.
Over time, it should become easier for researchers to separate the effects of AI from other forces. But they will still not be able to solve the question that politicians and everyday citizens want an answer to: What comes next?
“What the data can almost never tell us is where we will be in five to 10 years,” Ms. McEntarfer, the former Commissioner of Labor Statistics. “People are looking for data to answer that question, and it’s just too hard.”
