Fewer managers, more tech talent: How early AI investments rewired company org charts

The AI revolution sweeping through corporate America is reshaping who gets hired, who gets promoted, and which skills may become obsolete.

While the full impact on the workforce remains to be seen, new research on what happened when companies first started adopting artificial intelligence technologies may offer some clues. Between 2010 and 2018, as AI was starting to transition from research to commercial use, early adopters significantly reshaped their workforces, prioritizing STEM degrees and flattening out their management structures.

The paper, co-authored by Berkeley Haas professor Anastassia Fedyk, also provides the first tool for accurately tracking these changes at the level of individual companies.

“Historically, AI investment has been looked at on the aggregate scale—by industry or occupation—with some anecdotal evidence of one or another firm putting money into new programs,” says Fedyk, an assistant professor of finance. “Our work is the first comprehensive look across individual firms that measures their early investment in AI.”

A gold mine of employee résumés

Co-authored by Tania Babina and Alex He from the University of Maryland and James Hodson from the AI for Good Foundation, the paper is forthcoming as a chapter in the National Bureau of Economic Research (NBER) volume Technology, Productivity, and Economic Growth.

“Companies need three ingredients to make an AI strategy work: data, computing power, and human capital,” Fedyk says. “The lever that can be changed quickly and with a lot of impact is human capital.”

Given this, Fedyk and her colleagues analyzed data from hundreds of millions of résumés, representing 42% of all U.S. employees in 2010, when their sample started, and 63% of all employees by 2018, when their sample ended. This information essentially allowed the researchers to craft informal organization charts of many companies. Next, they paired this information with 180 million job postings in the U.S.

The challenge was discerning how likely it was that a particular hire would be working on AI-related issues. To do this, the research team connected four indisputably AI-related terms—artificial intelligence, machine learning, computer vision, and natural language processing—with tens of thousands of skills listed on job postings and résumés. A term like “deep learning,” for instance, is tightly correlated with AI; a term like “information retrieval” has a looser relationship but would not necessarily be tied to an AI-related job. Using these relationships allowed the researchers to determine which jobs, based on their descriptions, could serve as a proxy for investment in AI.

“Once we had that, we could simply count jobs at each company and see how that affected the rest of the workforce,” Fedyk says. “If a firm of 1,000 people hired five new AI employees, what happened to the other 995 people?”

Effects on workforce composition

At the basic level, based on prior work by Fedyk, they found that total employment went up: As AI-investing companies began to innovate more they created a greater need for new employees.

The composition of the workforce also changed in the wake of AI investments. First, the number of employees with a college education or beyond increased, while those with no college education declined—and not only because of the new AI hires. At the same time, these companies saw a general shift toward hiring workers with a degree in STEM fields and a reduction in new hires with social science degrees.

Finally, Fedyk and her colleagues found that AI-related hires made firms less top-heavy, defined by roles in middle-management and above. “We didn’t know which way this would go, but we found these hires were more likely to be independent contributors,” Fedyk says. “In the case of AI, companies could have these deputized workers who are smart, well-educated, and they don’t need as much oversight by middle management.”

Looking for causality

By themselves, these results don’t establish a causal relationship: It could be that investments in AI caused these changes in the workforce or that investment in AI signaled that the companies were doing well and could innovate more and hire more at the same time.

To more carefully parse the effect of firms’ AI investments, Fedyk and her colleagues compared companies that were essentially interchangeable but for one variable: their connection to universities. Some companies had pre-existing connections to universities like Carnegie Mellon or the University of Toronto—schools central to the development of AI research in the decades prior to corporate interest in the technologies. Other companies historically hired from schools that were equally strong in other ways (including overall research in computer science) but did not have particularly strong legacies in AI research.

“We were able to leverage which universities had these strong AI hiring networks before the commercial wave hit around 2010,” Fedyk says. “We found that, when this shock hit, the firms that had pre-existing connections to those AI-strong universities but were otherwise like their peers were able to hire AI workers and establish AI teams more easily. From those early AI teams flowed the other effects on workforce composition.”

With the same team of researchers and the same dataset on résumés and job postings, Fedyk has explored two other dimensions of AI’s influence on companies. One paper found firms that had invested early in AI were larger and had a greater market share—but these results pertained mostly to big companies. Another paper revealed how firms’ investment in AI makes them more pro-cyclical, meaning they earn especially high returns in market upswings but lose value during downturns. In the same paper, the researchers also found that AI-investing firms became more correlated with other growth firms and less correlated with value firms, consistent with AI increasing these firms innovation and creating growth options.

Together, these results help shed light on the early effects of AI. Since then, AI adoption has skyrocketed, with the latest wave of generative AI tools taking center stage. More work needs to be done to understand how those developments might change the landscape.

But so far, some of the early effects seem to be only strengthened by the new wave of AI: Big players with vast amounts of data and computing resources are poised to benefit most. Fedyk and co-authors found that early investments in AI were associated with increased industry concentration, benefitting the largest “superstar” firms. This may be an early warning sign, as concerns about increased concentration and market power have grown, she said.

Takeaways

Investing in AI tends to reshape a company’s workforce and organizational structure in several ways:

• Expanded employment: Firms typically increased hiring overall, indicating that AI-related innovation can stimulate growth rather than simply replacing jobs.
• Shift toward advanced education & STEM: New hires were more likely to have a college degree—often in technical fields. There were fewer hires with social science or medical backgrounds.
• Flattening of management: Companies became less “top-heavy,” suggesting that highly skilled independent contributors can function effectively with less managerial oversight.

Read the full paper:

Firm Investments in Artificial Intelligence Technologies and Changes in Workforce Composition
By Tania Babina, Anastassia Fedyk, Alex X. He, and James Hodson
National Bureau of Economic Research