At the ongoing 2026 Global Digital Economy Conference in Beijing, multiple leading enterprise representatives shared that artificial intelligence (AI) is rapidly integrating into all sectors, significantly unlocking corporate production potential. It is dramatically reducing production costs while enhancing quality and efficiency, poised to comprehensively reshape the entire social production system.
Wang Haihua, Deputy General Manager of the Beijing Humanoid Robot Innovation Center, stated that the explosion of large language models in recent years has allowed AI to complete its cognitive revolution. Today, embodied intelligence is propelling AI through an action revolution, enabling it to truly move out of virtual spaces and into the physical world, equipped with the tangible capabilities to perceive environments, adapt to scenarios, and operate autonomously.
JD.com Group Technology Committee Chairman and JD.com Cloud President Cao Peng indicated that currently, more and more enterprises are extensively utilizing new AI technologies across all stages including production, manufacturing, marketing, and customer service. "The JD.com Intelligent Agent Platform currently runs over 400 major AI applications, supporting nearly 40,000 intelligent agents, covering all business segments from procurement and sales to logistics, customer service, and R&D."
Xinmei Group, a digital marketing firm, saw its co-founder Wu Haifeng state that content production efficiency and quality directly determine the company's revenue capability. Previously, a three-person team would take three to five days to produce at most two advertising short videos. After introducing AI tools, one person can now produce over ten short videos per day. "As AI technology continues to mature, production efficiency and product quality will become increasingly higher, which is crucial for us."
The large-scale unleashing of AI productivity first depends on the capability leap and cost reduction of computing infrastructure. As large model iteration cycles shorten dramatically and intelligent agent applications rapidly proliferate, society-wide Token consumption is growing exponentially, testing the scale, stability, and cost-effectiveness of computing power.
Cao Peng explained that JD.com has built a full-stack AI product matrix from underlying computing power to upper-layer applications. Its systematic deployment has not only enhanced the stability of computing power supply but also significantly reduced the cost of using AI. Currently, internal weekly Token consumption at JD.com has reached the trillions level, with extreme cost optimization providing the economic foundation for AI's penetration across all business operations.
If computing power is the "power engine" for AI production, then intelligent agents are the "digital employees" for enterprises, driving corporate production systems from single-point tool applications towards fully-processed, self-driven intelligent upgrades.
Moonshot AI co-founder Zhang Yutao pointed out that current models already possess the capability to complete hour-level complex tasks. The core challenge has now shifted to constructing a closed-loop operating environment, allowing models to autonomously explore and solve problems within the cycle of "decision-making, execution, verification, and correction." He emphasized that as model capabilities improve, the "scaffolding" built for intelligent agents should be gradually removed, reducing the constraints of manual rules to fully unleash general capabilities. The ultimate value of intelligent agents lies in establishing human trust in them, enabling them to take on more complex production tasks.
Kingsoft Cloud Holdings Ltd (KC) Senior Vice President Liu Tao revealed that last year, the daily Token consumption per employee was still in the millions. Now it has grown to approximately 35 million, and employees outside the R&D system have also begun using AI programming to solve business problems. The proportion of AI-generated code within the company is increasing. Simultaneously, operational tasks such as update analysis, online alerts, and asset management are gradually being taken over by intelligent agents.
360 Group founder Zhou Hongyi predicts that this year, hundreds of millions of intelligent agents will be fully integrated into society, fundamentally reconstructing industrial logic, security rules, and governance systems.
Comments