Workers Hoard Expertise Amid AI Training Fears

More than one-third of workers report actively hoarding their expertise, driven by a fear that the artificial intelligence agents they are being asked to train will ultimately replace them. This trend highlights a growing tension between organizational demands for AI development and individual employee security.

Companies across various sectors are increasingly seeking to leverage AI for workflow automation, a process that necessitates access to real-world human operational data. Meta, for instance, began capturing employee mouse movements, clicks, and keystrokes on work computers in the spring of this year. This data collection was explicitly stated as a method to train its AI agents using authentic examples of human work processes. Similarly, McKinsey & Company developed its internal generative AI, Lilli, by drawing upon the extensive expertise accumulated within the firm over its long history. The director of design for Lilli articulated this strategy, explaining that for much of McKinsey's existence, "our knowledge was with our experts." Lilli represents a concerted effort to extract this specialized knowledge from individual experts and transform it into an organization-wide, accessible resource.

While this approach offers clear advantages from an organizational standpoint, enabling broader access to knowledge and potentially increasing efficiency, the benefits are less apparent to the individual expert. The act of codifying and transferring specialized knowledge to an AI system can be perceived as diminishing the unique value of the human expert, thereby increasing their vulnerability to job displacement. This concern is amplified by the ongoing wave of layoffs in the tech industry, with Meta itself initiating significant workforce reductions in May, shortly after the news of its click-tracking initiative became public. The layoffs, which affected approximately 8,000 employees, underscore the precariousness of employment in a rapidly evolving technological landscape, further fueling employee anxieties about job security in the face of advancing AI capabilities.

The phenomenon of expertise hoarding is a complex response to this perceived threat. Employees may intentionally limit the sharing of critical information, delay the documentation of their processes, or subtly resist the integration of AI tools that could automate their tasks. This behavior, while understandable from an individual's perspective, poses a significant challenge for organizations aiming to implement AI effectively. The success of AI initiatives often hinges on the quality and comprehensiveness of the data used for training, as well as the seamless integration of AI into existing workflows. When employees actively withhold knowledge, it can lead to incomplete or biased AI models, hinder the automation process, and ultimately undermine the intended benefits of AI adoption. This dynamic creates a critical need for organizations to address employee concerns proactively, fostering trust and transparency around AI implementation, and exploring strategies that emphasize human-AI collaboration rather than outright replacement.