IBM Executive: 2025 is 'Inflection Point Moment' for AI Agent Scale Adoption

A few days ago, Zhai Feng, general manager of technical sales and chief technology officer of IBM Greater China, published an article summarizing and predicting the development of AI Agent in 2025. In the article, Zhai Feng mentioned that 2025 will usher in the explosion of AI Agent. He explained that this judgment of the industry is closely related to the breakthroughs in technical fields such as large language model (LLM). It is reported that the core of the Agent lies in the autonomy and planning ability, they need to have logical reasoning, tool calling ability and the ability to perform tasks, and they need to realize high-speed, large-scale operation. In the past two years, more enterprises have begun to turn to small and medium-sized models with "faster speed, smaller demand for computing resources, stronger logical reasoning, contextual understanding, and external interaction capabilities," coupled with chain-of-thinking training, extended context windows, and function calling capabilities, Zhai Feng summarized the above information, saying that these key breakthroughs mean that the outbreak of AI intelligences in 2025 has already had a Technical foundation. Regarding the future of AI agents, Zhai Feng said, "The development has been overwhelming, but there is often a gap to be crossed between the ideal and the reality. Therefore, Zhai Feng mentioned that in order to build an AI Agent that autonomously handles complex decisions, it is not enough to rely only on the optimization of the algorithmic layer, and it is also necessary to make breakthroughs in contextual reasoning, edge case testing, and other aspects. In addition, Zhai Feng also emphasized that human beings must correct the tendency of "technology supremacy" with the idea of "human-centeredness": although the Agent has autonomous capabilities, it by no means means implies an all-round retreat of human beings. In this regard, Zhai Feng believes that Agent will play the role of "empowerment" in the future, and become a collaborator in the human-led intensive workflow.