Employees can save an average of 3% of their time by using AI chatbots, according to a new study.
If that doesn’t sound like a lot to you, the study agrees: It says that the AI-powered boosts in time saved and the wages earned all added up to deliver “no significant impact” for any occupation that was researched.
It’s a finding that hints at a potential collapse of the generative AI bubble, though likely not one that will comfort any of the hundreds of thousands of tech workers who have been laid off in recent years due to (at least in part) the hope of AI replacements.
3% More Time Saved and 3-7% Higher Pay
The research paper is out from the National Bureau of Economic Research, which examined the link between AI use and corporate records in Denmark.
The paper’s authors, assistant professor of economics at the University of Chicago Anders Humlum and economics PhD student at the University of Copenhagen Emilie Vestergaard, picked Denmark because the country’s AI adoption rate and hiring practices are comparable to the US, but it has better record-keeping. According to a Fortune article covering the report, the paper examined 25,000 workers across 7,000 workspaces within an impressive range of occupations, including software developers, IT support specialists, accountants, and many others.
This just in! View
the top business tech deals for 2025 👨💻
The big takeaway? AI chatbots did help workers, but not by much. In addition to the 3% time savings, employees saw between 3% and 7% increases in pay thanks to chatbot use.
“AI chatbots are now widespread—most employers encourage their use, many deploy in-house models, and training initiatives are common. These firm-led investments boost adoption, narrow demographic gaps in take-up, enhance workplace utility, and create new job tasks. Yet, despite substantial investments, economic impacts remain minimal.” -the abstract
Perhaps AI Can Replace 3% of a Worker?
A 3% time savings might seem to indicate a benefit when stretched out to an international corporation. However, the remaining 97% of specialized work that each workers contributes is less likely to be replicatable by a chatbot: It might be able to speed up the writing of a thank you email, but it can’t catch a math error in an accountant’s monthly bank reconciliation.
In short, the study serves as a strong indicator that AI chatbots are not able to save a worker so much time that a business can reduce its headcount as a result.
Is Generative AI a Bubble?
Over the past couple of years, AI has been an incredibly quickly adopted technology, earning untold billions of dollars in investments while driving news cycles on everything from government regulation to college cheating scandals. Now, agentic AI, the next evolution of the chatbot, is getting tech executives everywhere excited about savings potential.
However, part of the downside of that speedy rollout is that we have yet to see AI tools fully deliver on the many promises that have helped spur the rapid adoption. Now, large-scale studies like the National Bureau of Economic Research are emerging, and indicating that AI chatbots cannot live up to the hype.
Perhaps agentic AI will prove to offer a far greater value than chatbots. Until then, though, we may want to keep 100% of our human staffers around.
