Instead of pushing down wages or working, generative AI chatbots like ChatGpt, Claude and Gemini have had little or no significant wage or labor impacts to date.
In a working paper released earlier this month, economists Anders Humram and Emily Vestergaard saw the impact of AI chatbots' labor market on 11 occupations covering 25,000 workers and 7,000 workplaces in 2023 and 2024.
Many of these professions are said to be able to support accountants, customer support specialists, financial advisors, HR specialists, IT specialists, journalists, legal experts, marketing experts, office clerks, software developers and teachers.
However, after Humlum, Vestergaard, an assistant professor of economics at the Booth Business School at the University of Chicago and a doctoral student at the University of Copenhagen, analyzed the data and found that the impact of chatbot labor and wages was minimal.
AI chatbots have no significant impact on revenue or recorded time in any occupation
“AI chatbots have not had a significant impact on revenue or recorded time in any occupation,” the author states in his paper.
The report should concern a high-tech industry that has cultivated infrastructure to support billions while hyping the economic potential of AI. Earlier this year, Openai admitted that even the most expensive enterprise SKUs cut their money with each question, but companies like Microsoft and Amazon are beginning to pull back AI infrastructure spending in light of low business recruitment past a few pilots.
The problem isn't that workers are avoiding generative AI chatbots – the exact opposite. But they simply do not equate with actual economic benefits.
The adoption of these chatbots was very fast…but looking at the economic outcomes, it really doesn't move the needle
“The adoption of these chatbots is very fast,” Humlum said. Register. “Most workers in exposed occupations now employ these chatbots. Employers are also shifting gear and actively encouraging it. But looking at the financial outcomes, they don't actually move the needle.”
Researchers looked at how corporate investment in AI contributed to the recruitment of AI tools workers, and how chatbot recruitment affected the workplace process.
Enterprise-led investment in AI has encouraged the adoption of AI tools – saving 64-90% of users across the occupations surveyed, but chatbots have had a complex impact on job quality and satisfaction.
The economists, for example, found that “AI chatbots created new job tasks for 8.4% of workers, including those who were not using the tool.”
In other words, AI is creating new tasks that cancel some potential time savings by using AI in the first place.
“One of the very harsh examples of being close to home for me is that there are a lot of teachers who say they are spending their time trying to detect whether students are using ChatGpt by tricking their homework,” Humlum explained.
He also observed that many workers say they are currently spending time reviewing the quality of AI output and write prompts.
Humlum argues that automation tools historically tend to generate more demand for workers on other tasks, or more positively and negatively spin as a subtraction from potential productivity gains or more positively and negatively.
“These new employment tasks create new demand for workers, which could raise wages if these are higher additional tasks,” he said.
Overall, however, using AI has saved less time than expected. According to the survey, “users report an average time saving of just 2.8% of working hours” through the use of AI tools. This is more than an hour per 40 hours a week.
The authors note that this finding differs from other randomized controlled trials that found productivity benefits on the order of 15%. And they explain this contradiction by saying that other research focuses on professions where AI is likely to be productive, and that real-world workers are not operating under the same conditions.
“So I think there are two important reasons why real economic benefits are lower. [the cited studies]”Humlum notes that his research relies on actual tax data.
“Firstly, most tasks don't fall into the category where ChatGpt can automate everything, and secondly, they are in this middle stage where employers are still awakening to a new reality, trying to figure out how to best realize the possibilities of these tools.
If there is a productivity boost, Humlum and Vestergaard estimate that a small portion of the benefits (3-7%) will be passed on to workers in the form of higher revenue.
Humlum said, “There is no doubt that there is a question of who they are actually accumulating and it can be a company. We cannot see the profitability of a company directly. We can save time on existing tasks, but we cannot scale the output.
“So it's like saving time writing emails. But if you can't actually do more work or do other things that are really valuable, it will put the damper on how much time you should actually expect in your revenue ability, total time, and wages.”
Humlum said the impact of using AI chatbots can be improved through the company's commitment to internal education and evangelism in the form of productivity, time savings and work quality. He particularly pointed out how solid initiatives can reduce the gender gap with tools. Few women use these tools than men.
But doing so at this point shows little promise in return.
“When we are looking at hard metrics in terms of economic outcomes — in the administrative labor market data on revenue, these tools really didn't make a difference,” Humlum said. “So, at least in the short term, I think we'll put a cap in some way about what returns we should expect from these tools.
“My general conclusion is that any story you want to tell about these tools being so transformative, they need to be fought at least two years later. [the introduction of AI chatbots]they are not making a difference in their financial outcomes. ”®
