Meta Chief Executive Mark Zuckerberg admitted during an internal meeting on Thursday that the company’s AI agent systems have advanced more slowly than anticipated, according to a recording obtained by Reuters. He also noted that a recent reorganization involving significant staff reductions was less orderly than planned and that timing estimates had been off. Zuckerberg and other leaders have sought to adjust some of the earlier organizational shifts without altering the overall direction. In May the company reduced its global workforce by about 10 percent and moved roughly 7,000 employees to AI-related roles, actions that drew internal resistance and raised morale issues. The restructuring was intended to support heavy spending on artificial intelligence infrastructure and to capture efficiency improvements from AI-supported operations. In May, Zuckerberg told staff he did not foresee additional broad layoffs this year, though some employees remained doubtful. Looking back, he observed that progress on agent development over the past four months had not sped up as hoped and that expectations for the new structure had not yet been met. He was referring to AI agents, which are automated tools designed to perform tasks for users. During planning discussions in January and February, executives had expressed concern about moving too slowly, he said, and were highly optimistic about tools such as Claude Code from Anthropic. Meta is expected to invest up to $145 billion in AI infrastructure this year, representing a large share of the more than $700 billion that major technology firms plan to spend overall. Zuckerberg stated he anticipates more substantial returns from these investments within the next three to six months. A company spokesperson declined to comment. In the same meeting, Chief Technology Officer Andrew Bosworth reported that an internal review of a data security issue involving the firm’s mouse-tracking software found no employee data had been used for AI training. The program, which monitors mouse movements and digital activity, was paused last month. If resumed, participation will be voluntary, Bosworth said. When the software was first deployed on U.S. employee computers in April, there was initially no option to decline.

Credit:
https://indianexpress.com/article/technology/tech-news-technology/metas-zuckerberg-says-ai-agent-tech-progressing-slower-than-expected-10769142/
BCN