Meta Platforms has started rolling out internal tracking software on devices used by its U.S.-based employees, gathering data such as mouse movements and keystrokes to support the training of artificial intelligence systems.

According to reports, the rollout is part of a wider shift in how the company structures its internal operations, with a growing focus on AI systems designed to handle routine workplace functions.

Employees are now being encouraged to rely more heavily on AI tools for tasks like coding, even in situations where it may slow productivity in the short term. The company has also introduced a more general “AI builder” role and reshaped parts of its engineering teams to align with AI-driven development.

As part of this transition, Meta has set up a dedicated Applied AI engineering group tasked with building systems that can write and deploy software. Some engineers have already been reassigned to support this effort.

The internal tool, known as Model Ability Initiative, or MCI, operates across work-related applications and websites. An internal memo shared within Meta SuperIntelligence Labs noted that the system can also capture periodic screenshots from employee screens.

Company documents indicate that the collected data is being used to improve how AI systems understand real-world computer use. The focus is on everyday interactions, including keyboard shortcuts, menu navigation, and managing different application states. One internal note described it as a way for employees to help improve models “simply by doing their daily work.”

Meta’s Chief Technology Officer Andrew Bosworth said in a separate internal message that the company is increasing data collection under its “AI for Work” initiative, now renamed the Agent Transformation Accelerator.

“The vision we are building towards is one where our agents primarily do the work and our role is to direct, review and help them improve,” Bosworth said. He added that these systems are being designed to identify moments where human input is needed so they can improve over time.

Bosworth noted that Meta is focusing on building datasets and evaluation frameworks based on employee workflows. While he did not detail the training process, he said workplace interaction data would remain central to development.

Meta spokesperson Andy Stone said the information collected through MCI is used only to train models and not to assess employee performance. “If we’re building agents to help people complete everyday tasks using computers, our models need real examples of how people actually use them,” he said, pointing to actions such as clicking buttons and navigating menus.

The company added that safeguards are in place to avoid collecting sensitive data, though it did not specify what types of information are excluded.

Meta tests AI-powered Zuckerberg and expands social commerce strategy

Alongside these efforts, Meta is also exploring new ways to use AI for internal communication. The company is developing a photorealistic digital version of its chief executive, Mark Zuckerberg, designed to interact with employees in real time.

The system is being trained on Zuckerberg’s voice, tone, and communication style, allowing it to simulate natural conversations and deliver responses aligned with his views. The goal is to give staff more direct access to leadership, particularly across a global workforce.

Although still in early development, the project highlights Meta’s continued investment in virtual human systems capable of speaking and responding across different settings.

At the same time, Meta is strengthening its position in social commerce by linking creators, AI tools, and advertising more closely to shopping activity on platforms such as Instagram and Reels. The update is expected to make it easier for users to move from discovering content to completing purchases within the same experience.

Go to Source to See Full Article
Author: Rony Roy