Research Roundup: AI’s Uneven Impact on Work

Generative AI is spreading through workplaces, but its impact is far from uniform.

Some employees take on additional work to develop AI solutions. Others have yet to adopt the technology, even when colleagues perform similar tasks. Meanwhile, younger workers in AI-exposed occupations face different employment trends from their older peers.

For business schools preparing future managers and graduates, these differences matter. Four recent working papers explore the hidden work behind AI innovation, uneven adoption, and emerging patterns in entry-level employment.

The Hidden Work Behind AI Innovation

AI promises to save time. But what happens when employees must spend additional hours making it work?

Researchers studied employees developing AI solutions at an academic medical center and a law firm. Both organizations were trying to turn individual experiments into solutions that could be used more widely.

At the law firm, employees discovered that AI-generated legal research could look convincing yet contain fabricated citations. 

Hauntingly beautiful results.  
“The output looked beautiful, but half the citations were phantoms. —Interview excerpt

Correcting such errors was only part of the challenge. Developing reliable AI solutions required employees to experiment with prompts, document what worked, review results with colleagues, and repeatedly adapt their approaches as the technology changed. These activities often added to existing responsibilities rather than replacing them.

The researchers call this tendency experimentalist work intensification. Both organizations experienced it, but employees at the law firm reported greater strain. At the medical center, workers received more structured support for experimentation, while the law firm’s employees faced greater difficulties obtaining assistance and recognition for their additional work.

The study suggests that managers need to organize, support, and recognize the work involved in developing AI, not simply provide access to the technology.

Authors: Arvind Karunakaran, Stanford University; Katherine Kellogg, Massachusetts Institute of Technology; and Batia Mishan Wiesenfeld, New York University

Read the research   ↗

AI Is Everywhere. So Why Isn’t Everyone Using It?

Generative AI has reached a wide range of professions, but that doesn’t mean most workers are using it.

Researchers found that more than 80 percent of occupational categories have AI adoption rates above 20 percent. Yet within most occupations, fewer than half of workers actually use the technology.

Differences also emerge among people performing the same tasks. For example, while 61 percent of workers preparing research reports use AI for that activity, nearly four in 10 do not. Similarly, among workers analyzing data to identify trends, 58 percent use AI, while 42 percent do not.

The researchers describe this pattern as widespread but shallow adoption. Their analysis also suggests that experience with AI may help workers extend its use to other activities.

The findings highlight the difference between AI’s potential and its actual use. A job may include many tasks that AI could assist with, but that doesn’t mean the people doing that job are using it.

Authors: Alexander Bick, Federal Reserve Bank of St. Louis; Adam Blandin, Vanderbilt University; David J. Deming, Harvard University; and Tyler R. Schumacher, Vanderbilt University

Read the research   ↗

AI Can Make Work Faster—but Not Every Task Is a Good Fit

Generative AI can save consultants time, but its value changes depending on what they ask it to do.

In interviews with 33 consultants from leading German consulting firms, 70 percent reported efficiency gains from using generative AI. The technology was particularly useful for activities such as producing first drafts, synthesizing information, and handling administrative work, freeing consultants to spend more time on strategic and client-facing tasks.

“But those gains came with limits.

But those gains came with limits. Three-quarters of participants emphasized the need to cross-check AI-generated information, while 70 percent preferred traditional research methods for specialized topics. The researchers argue that effective use therefore depends on what they call “Task-GenAI Fit”: matching the technology to the demands of a particular task rather than assuming it will improve every type of work.

For managers, the implication is straightforward: Deciding where to use AI may matter as much as deciding whether to use it at all.

Authors: Matthias Tuczek and Michael H. Breitner, Leibniz University Hannover; Kenan Degirmenci and Kevin C. Desouza, Queensland University of Technology; Richard T. Watson, University of Georgia

Read the research   ↗ 

Is AI Changing the First Step on the Career Ladder?

Employment among 22–25-year-olds fell 11 percent in occupations with higher AI exposure, while rising 10 percent in those with lower exposure. The gap was largely driven by reduced hiring, raising questions about how graduates enter the workforce.

Using U.S. payroll records covering millions of workers through June 2026, researchers examined how employment trends varied by age and occupational AI exposure.

They found no widespread employment decline across AI-exposed occupations. Instead, the divergence was concentrated among younger workers, particularly in occupations where AI is used to automate tasks rather than assist employees. Older workers showed no comparable employment gap.

The findings suggest that entry-level opportunities may be changing, but the data alone does not establish that AI caused the discrepancy. Educational differences and employment trends that began before generative AI became widespread may also help explain the decline. 

Authors: Erik Brynjolfsson, Bharat Chandar, and Ruyu Chen, Stanford University

Read the research   ↗

 
Together, these studies show that AI’s true impact on work is shaped by several factors: how organizations develop solutions, how employees adopt tools, and how employment opportunities evolve. For business schools, the findings raise important questions about effectively preparing graduates to navigate and manage these changes.