New York: Meta Chief Executive Mark Zuckerberg has admitted that the company’s work on artificial intelligence agents is moving more slowly than expected, highlighting the difficulties technology companies face in turning powerful AI systems into reliable digital assistants that can complete complex tasks on their own.
Speaking during an internal company meeting, Zuckerberg told employees that progress in AI agent development over the past four months had not accelerated as quickly as Meta had hoped. However, he remained confident about the company’s direction and said that clearer results from its massive investments in artificial intelligence could become visible within the next three to six months.
The comments offer a rare glimpse into the challenges facing Meta as it spends heavily to compete in the global AI race. The company has invested large amounts of money in advanced computer chips, data centres, research teams and highly skilled AI specialists. It has also reorganised its artificial intelligence operations in an effort to speed up development and compete more strongly with leading companies in the field.
AI agents are considered an important next step in the development of artificial intelligence. Unlike ordinary chatbots that mainly respond to questions or create text, AI agents are expected to perform several connected tasks, use digital tools, make decisions and complete work with limited human supervision.
However, developing reliable AI agents has proved difficult across the technology industry. While AI models can perform well in controlled tests, they can still make mistakes when handling longer and more complicated tasks. They may lose track of instructions, make poor decisions or fail when they face unexpected situations.
Meta’s difficulties come at a time when the company is making major changes to its AI operations. The company has moved thousands of employees towards artificial intelligence related work and has been recruiting researchers and engineers with specialised skills. These changes have increased pressure inside the organisation as teams work to produce faster results.
Despite the slower progress in AI agents, Meta continues to push forward with its wider AI ambitions. The company has placed increasing attention on the idea of personal superintelligence, a vision in which highly advanced AI systems could help individuals with communication, creativity, learning and everyday tasks.
There are also signs that Meta may be making progress in developing stronger AI models. Recent reports have suggested that company executives believe some internal models are becoming more competitive with systems developed by leading AI companies. However, such internal performance claims have not been independently confirmed.
The company is also paying closer attention to the safety and security of AI agents. As these systems become more independent, researchers are increasingly concerned about how they use tools, access information and make decisions. A powerful AI agent that acts incorrectly could create greater risks than a chatbot that only provides a written response.
Another issue raised in recent reports involves the use of employee computer activity for AI training purposes. Meta reportedly paused a programme connected to collecting such information after concerns were raised. An internal review found that employee data had not been leaked, and the company was considering whether the programme could return on an optional basis.
Zuckerberg’s remarks show that even companies with enormous financial resources and access to advanced technology are finding it difficult to create dependable AI agents. Building a powerful language model is only one part of the challenge. Making that model understand a goal, plan several steps, use tools correctly and recover from mistakes is a much more complicated task.
The situation also reflects a wider reality across the technology industry. Companies have promoted AI agents as systems that could eventually help people manage schedules, conduct research, handle office work, shop online and complete other complicated digital tasks. Yet the technology still faces problems with accuracy, reliability and the ability to work independently for long periods.
For Meta, the coming months could be crucial. The company has made artificial intelligence one of its biggest priorities and has committed enormous resources to the field. Investors, employees and competitors will be watching closely to see whether those investments begin producing the results Zuckerberg expects.
The slower than expected development does not mean Meta is stepping away from its AI plans. Instead, it shows the scale of the challenge ahead. The race to build useful and dependable AI agents is continuing, but Zuckerberg’s comments suggest that the road towards truly independent digital assistants may be longer and more difficult than the technology industry once expected.