AI helps employees learn faster but hits an ‘expertise wall’: study

MASSACHUSETTS, UNITED STATES — Generative artificial intelligence (AI) can help employees upskill quickly, but it may not turn novices into experts, according to new research that underscores limits in the future of work. 

A study by Stanford University and Harvard Business School found that while AI can boost performance on certain tasks, employees still rely heavily on domain knowledge to produce expert-level results.

AI narrows gaps but can’t replace expertise

The study involved 78 employees at IG Group, a United Kingdom-based fintech firm, divided into three groups: experienced writers, marketing specialists with some domain knowledge, and technology specialists with no writing background. 

Participants were asked to conceptualize and write articles, with some receiving AI assistance.

Without AI, expert writers outperformed non-experts in both tasks. With AI, the gap narrowed for conceptualization.

“Marketers using AI slightly outperformed writers using AI—and all three groups that used AI outperformed writers who didn’t,” the researchers reported. However, AI did not fully close the gap in writing quality. 

Experts still led, with marketing specialists aided by AI close behind, while technology specialists’ scores remained largely unchanged.

Researchers described this limit as the “AI wall,” a boundary beyond which AI cannot compensate for lack of expertise. 

“Nonexperts using AI did better at conceptualization because it required less expertise than writing did,” the study noted. Writing demands nuanced judgment and domain-specific skills, which technology specialists lacked.

Implications for workplaces of tomorrow

The findings highlight critical considerations for organizations integrating AI into workflows.

“AI isn’t a magic fix for everything at work if it is not able to fully automate tasks. When AI can’t do the job alone and it replaces experts, it will help some people narrow the gap… but only in certain situations and when the conditions are right. It’s not a one-size-fits-all solution,” Luca Vendraminelli, the Stanford postdoctoral researcher who led the study, emphasized.

Experts enable more effective collaboration with AI. Marketing specialists, familiar with writers’ language, could refine AI-generated content, whereas technology specialists could not. 

For businesses, this signals that deploying AI requires rethinking roles, processes, and team structures. 

Job redesign may be necessary to take full advantage of AI’s potential, particularly in areas like content strategy, marketing, and cross-functional projects.

As AI reshapes work, expertise remains irreplaceable. “AI can only take people so far,” Vendraminelli said. 

“Expertise is irreplicable. No technology can substitute for it,” Vendraminelli added.

The study suggests that the future of work will increasingly combine human judgment with AI assistance, rather than rely on AI alone to deliver expert outcomes.