Mark Zuckerberg, CEO of Meta, has admitted that the development of artificial intelligence (AI) agents within the company is not progressing as quickly as management had hoped, despite the massive investments and extensive restructuring plans, including layoffs and the creation of AI-focused teams.

Massive Investments, But Less Than Expected Results

Meta continues to increase its spending on AI infrastructure while cutting costs in other areas. This includes laying off around 8,000 employees last May and redistributing about 7,000 others into teams focused primarily on AI technologies. However, during an internal meeting, Zuckerberg acknowledged that progress in developing "smart agents" has not accelerated as the company had hoped in recent months.

The CEO of Meta stated that the development of these technologies over the past four months has not accelerated as expected, noting that the company is still on a long journey towards achieving what he describes as "superintelligence."

What are AI agents?

AI agents or AI Agents are systems capable of performing tasks almost independently based on artificial intelligence models, such as searching for information, managing tasks, and interacting with digital applications and services on behalf of the user. This technology is one of the most prominent future bets for major tech companies, which seek to transform artificial intelligence from a mere conversational tool into a digital assistant capable of taking action and completing work directly.

Meta spends $145 billion on AI

According to reports, Meta's spending on artificial intelligence infrastructure is expected to reach around $145 billion this year in a fierce race with companies like OpenAI, Anthropic, and Google. Zuckerberg confirmed that the company expects to start seeing more visible results of its massive investments within three to six months.
Meta had previously acquired Moltbook, which was famous for developing applications and AI agents, and integrated its team into the company's artificial intelligence projects.

Zuckerberg revealed that several senior executives within the company had been concerned since the beginning of the year about the slow pace of restructuring and the shift towards artificial intelligence. He explained that some leaders felt the company was not moving fast enough compared to its competitors, especially with the increasing success of tools like Anthropic's Claude Code.

Last March, Meta established a new division called "Applied AI" and transferred around 6,500 engineers and product managers from different teams to work on artificial intelligence projects. However, these changes did not gain full acceptance within the company. In addition to laying off about 10% of its global workforce, Meta redirected thousands of employees to artificial intelligence teams and canceled nearly 6,000 positions it had planned to hire for in the future. Andrew Bosworth, Chief Technology Officer at Meta, acknowledged that the company failed to adequately explain the goals of these changes to employees, describing the internal communication about restructuring as "very bad."

Meta Backs Away from Tracking Employees for AI Training

In a related context, Meta faced widespread criticism after launching a program that relied on tracking employees' mouse movements and keyboard presses to gather data to help train artificial intelligence agents. Concerns about privacy led to the temporary suspension of the program, before the company later confirmed that internal reviews had shown that employee data was not used in model training. The management also clarified that participation in any similar programs in the future would be entirely optional and no employee would be forced to join.