More than a third of workers say they’re hoarding expertise because they fear being replaced by the AI agents they’re being asked to train

Meta, a leading tech company, recently informed thousands of U.S. employees that their work computers would begin tracking mouse movements, clicks, and keystrokes. The purpose of this data collection was to train AI agents on real-world examples of human work. However, this is not an isolated case. Companies across the industry are seeking access to employees' expertise to train AI agents and automate workflows.

McKinsey, for instance, developed its internal generative AI, Lilli, by leveraging the knowledge and expertise accumulated over the firm's long history.

The director of design for Lilli explained that while most of McKinsey's history has been built on the knowledge of its experts, Lilli aims to disseminate this expertise throughout the organization. From the company's perspective, this makes perfect sense. However, the situation is less clear from the perspective of the experts involved.

In May, after the Meta story emerged, the company announced layoffs of 8,000 employees, citing AI as the reason. More broadly, over 175,000 tech workers have been laid off in 2026, with AI cited as the primary cause.

The situation is not unique to Meta or McKinsey. Companies worldwide are increasingly relying on AI agents, and they need access to human expertise to make these agents effective. This creates a dilemma for employees: they are asked to transfer their knowledge to AI systems, yet they fear being replaced by those very systems. In fact, a recent survey of 4,000 workers found that 35% are actively hoarding knowledge to avoid being replaced by AI, while 38% are hesitant to train colleagues in areas they consider their personal strengths.

Knowledge transfer is not a straightforward process. While some expertise can be easily captured, such as templates, playbooks, and workflows, other forms of knowledge are more challenging to extract. For example, an experienced salesperson's hesitation in a client conversation may not be easily observable or translatable into data.

Similarly, the alternatives an expert considered and rejected, the exceptions they have learned to recognize, and the intuitions they have developed through experience may not be easily identifiable through mere observation.

Companies are tempted to coerce employees into participating in knowledge transfer efforts, but this approach is unlikely to succeed. Compliance can be performative, meaning that employees may attend sessions and answer questions without truly transferring the valuable knowledge they possess. In fact, a recent study found that 35% of workers intentionally hoard knowledge specifically because they fear being replaced by AI.

This reluctance is often referred to as "evasive hiding," where employees attend sessions and comply with requests, but still withhold critical information.

To overcome these challenges, companies must approach knowledge transfer differently. They need to understand the nuances of knowledge hiding and create an environment where employees feel comfortable sharing their expertise. This requires open communication, trust, and a recognition of the value that experts bring to the table. By fostering a culture of cooperation and collaboration, companies can unlock the full potential of their employees' knowledge and overcome the limitations imposed by AI.