At the 2026 Lujiazui Forum, Gu Shu, Chairman and Executive Director of Agricultural Bank of China, participated in a keynote speech themed "Technological Innovation Empowers High-Quality Financial Development."
Addressing the application of AI agents in finance, Gu Shu noted their widespread use and active exploration within the industry. However, he pointed out that the definition of what constitutes an "AI agent" still requires further standardization and unification across the sector. For instance, when banks report their AI application achievements, the number of agents developed varies dramatically—some cite hundreds, while others claim tens of thousands. This discrepancy stems largely from the lack of a unified definition and standard.
He further elaborated that the ultimate count of agents deployed is less critical than the tangible business outcomes they deliver. Nevertheless, in the current phase of promoting and applying AI agents, it is crucial for every participant in the industry to identify their specific role and learn from each other's strengths. Without relatively standardized criteria, it becomes difficult to define what qualifies as a mature AI agent application. In the absence of such standards, accurately assessing an institution's relative standing within the industry and effectively leveraging others' best practices to collectively elevate the sector's application level becomes a significant challenge.
"At its core, an AI agent is an intermediary, proxy, or executor that enables large language models to function, and there is considerable variation in their development processes," Gu Shu stated. On this point, he proposed a recommendation: when designing agents to perform tasks, could those with relatively singular functions and fixed business processes be developed as "standardized components"? The advantage of creating such "standardized components" lies in avoiding redundant development and enabling their repeated reuse. This approach would allow the industry to focus more energy and resources on developing agents with autonomous planning and decision-making capabilities, which he described as a more meaningful and efficient endeavor.
Comments