Generative AI dramatically speeds up brainstorming and planning, cutting conceptualization time by nearly two-thirds, but a new Harvard Business School study reveals a hard limit: the technology cannot transform novices into domain experts, especially when execution requires deep contextual knowledge. The finding challenges the popular assumption that AI can democratize expertise across organizations. While AI makes employees feel capable of tackling unfamiliar work, the reality is more nuanced. Researchers at Harvard Business School and Stanford University conducted an experiment with 78 employees at IG Group, a global derivatives trading firm, to understand where AI truly excels and where human expertise remains irreplaceable. Where Does Generative AI Actually Help Most? The study organized participants into three groups: 12 web analysts (the domain experts), 26 marketing specialists (adjacent to the task), and 40 technology specialists like software developers (distant from the task). All groups were asked to write investing articles for the company's website using the same generative AI model. The results showed a striking pattern. When participants focused on organizing ideas and creating outlines, AI leveled the playing field. Technology specialists, marketing specialists, and web analysts all scored around 4.1 out of 5 on conceptualization tasks. But when execution began, the gap widened dramatically. Articles written by technology specialists with AI assistance scored 3.42 out of 5, roughly 13% below the 3.96 average for web analysts and 3.92 for marketing specialists. "AI makes you feel like you can do anything. But can you do a task as well as people whose job it is? That's what we set out to find out," said Iavor Bojinov, Associate Professor at Harvard Business School. Iavor Bojinov, Associate Professor, Harvard Business School The productivity gains, however, were substantial across all groups. Conceptualization took 23 minutes with AI compared to 63 minutes without it, a reduction of nearly two-thirds. Writing took 22 minutes with AI versus 87 minutes without it, a reduction of nearly three-quarters. Why Can't AI Bridge the Expertise Gap? The answer lies in what researchers call "knowledge distance." Think of it like learning musical instruments: someone trained on one woodwind instrument can pick up another woodwind more easily than switching to a string instrument. The closer your existing expertise is to the task, the more effectively you can use AI's suggestions. Marketing specialists and web analysts both had similar expertise profiles rooted in content creation and audience understanding. When AI suggested information or phrasing, they could evaluate it critically, spot gaps, and refine it based on their lived experience. Technology specialists, by contrast, lacked the contextual intuition to interpret and apply AI's advice effectively. "The folks who are too far away from the domain experts lack either sufficient understanding of the necessary information or lack the skills to use it effectively and therefore cannot match the expert's level of quality," explained Edward McFowland III, Assistant Professor at Harvard Business School. Edward McFowland III, Assistant Professor, Harvard Business School McFowland noted that conceptualization benefits from a structured, template-like approach that technical specialists naturally follow. Writing, however, demands creative judgment and nuanced decision-making rooted in domain experience. AI can provide the map, but navigating the terrain requires lived experience. How Organizations Can Leverage These Findings - Ideation and Planning: Use AI to accelerate brainstorming, outlining, and project framing across all skill levels, since AI performs equally well for these structured tasks regardless of user expertise. - Adjacent Role Transitions: Deploy AI to help employees move into roles similar to their current expertise, such as moving a data scientist into a marketing analyst position with significantly reduced retraining time. - Organizational Flattening: Shorten learning curves for new tasks like SEO optimization or content strategy by pairing AI with employees who have foundational domain knowledge in related areas. - Execution Caution: Recognize that AI cannot substitute for expertise in complex execution tasks; pair AI tools with experienced practitioners rather than expecting novices to match expert output. The research suggests that companies should rethink how they deploy AI in hiring and role design. Rather than assuming AI eliminates the need for expertise, organizations should use it to extend the reach of skilled workers and accelerate learning for people with adjacent skills. Bojinov emphasized the organizational implications: "You can have flatter organizations where the learning curves for new tasks, such as SEO optimization, become much shorter and almost disappear." However, this applies most effectively when employees already possess foundational knowledge in a related domain. Bojinov The study, titled "The GenAI Wall Effect: Examining the Limits to Horizontal Expertise Transfer Between Occupational Insiders and Outsiders," was conducted by researchers from Harvard Business School and Stanford University in 2024. The findings offer a reality check for business leaders betting on AI to solve talent shortages: the technology is a powerful productivity multiplier for ideation and a capable assistant for adjacent skills, but it cannot compress years of domain expertise into weeks of AI-assisted work.
