Meta tracks employee clicks and keystrokes to train AI agents, raising fresh fears over the future of tech jobs

Meta Platforms, Inc. has begun tracking how employees interact with workplace computers, including mouse movements and keystrokes, as part of a new internal effort to train advanced artificial intelligence systems. The initiative has sparked renewed debate over workplace surveillance, automation, and whether AI is moving from assistant to replacement inside the technology industry.

According to reports, the program is designed to study how people complete everyday digital tasks. By observing actions such as selecting menus, using shortcuts, navigating software, and managing workflows, the company aims to build AI agents that can imitate real human behaviour on computers. The move signals a significant step in how major companies are preparing for a future where software may handle a growing share of office work.

The internal system, reportedly called the Model Capability Initiative, is intended to help AI learn directly from real workplace behaviour rather than relying only on written instructions or static datasets. That means the systems can study how experienced workers solve problems, switch between tools, and complete tasks efficiently.

This method could allow AI agents to understand not just what to do, but how people actually do it in practical settings. For example, instead of being told to prepare a spreadsheet or organise information, an AI system could observe how an employee opens programs, sorts files, compares documents, and completes the assignment step by step.

Executives have described a long term vision in which AI agents carry out more operational work, while human staff supervise results, guide strategy, and improve system performance. That language has drawn strong attention because it suggests a future where many routine digital roles may be redefined.

For years, AI tools were marketed as assistants that could save time by drafting emails, summarising meetings, or helping write code. What is changing now is the level of autonomy being pursued.

Instead of merely responding to prompts, newer AI agents are being built to take action. They can manage sequences of tasks, make choices based on instructions, and continue working with limited oversight. If trained on enough real workplace behaviour, such systems could become increasingly capable of replacing repetitive knowledge work.

That shift matters because many white collar roles depend on exactly those processes: data entry, reporting, scheduling, research support, coding assistance, documentation, and internal coordination. If AI becomes reliable at handling these responsibilities, companies may decide fewer employees are needed.

The timing of this development has amplified concern. Meta Platforms, Inc. has already faced scrutiny over workforce reductions and restructuring efforts in recent years. Reports indicating further cuts or internal role changes have led many observers to connect efficiency drives with expanding AI investment.

The company has also introduced broader AI focused responsibilities, showing that traditional job categories may be evolving. Rather than hiring only for narrow specialist positions, firms increasingly want workers who can manage AI systems, automate workflows, and deliver more output with fewer resources.

This pattern is not limited to one company. Amazon.com, Inc. and other major firms have also reduced thousands of positions while increasing spending on automation, cloud infrastructure, and generative AI tools.

For workers, the immediate impact may not always be direct replacement. In many cases, AI first changes the nature of a job before it removes one. Roles may become smaller, more demanding, or more focused on oversight rather than execution.

Employees may be expected to produce more work using AI systems, manage multiple tools, or shift into higher value tasks that machines cannot easily replicate. Those who adapt quickly may benefit. Those in highly repetitive digital roles could face growing pressure.

The larger concern is transparency. When companies collect behavioural workplace data to train AI, employees may ask where the boundaries are. How much monitoring is acceptable? How is the data stored? Who can access it? And will workers be informed when their own habits are used to build systems that may later reduce staffing needs?

The debate around Meta Platforms, Inc. reflects a broader turning point for the modern workplace. AI is no longer only about chatbots or creative tools. It is becoming operational software trained to perform jobs once considered secure.

That does not mean human work disappears overnight. Judgment, leadership, relationship building, ethics, and complex decision making remain areas where people are essential. But the structure of many office roles could change faster than expected.

Companies see opportunity in speed and lower costs. Workers see uncertainty around stability and long term relevance. Governments and regulators may soon face pressure to address privacy, monitoring, and labour impacts as these systems spread.

The latest move is more than an internal experiment. It is a signal of where the technology sector may be heading next. If one of the world’s biggest digital companies believes employee behaviour is valuable training data for AI workers, others may follow.

That could accelerate a global race to automate administrative and technical tasks across industries such as finance, customer support, healthcare operations, logistics, and media.

For now, one message is clear: the question is no longer whether AI will sit beside workers. It is whether AI is being trained to sit in their place.