The 2026 World Artificial Intelligence Conference (WAIC) entered its second day on July 18th, a Saturday, drawing massive crowds. A key observation from the event floor was a significant shift in focus. Unlike previous years where the industry concentrated on comparing large model parameters and conversational abilities, this year's core theme was AI Agents. From personal assistants for consumers to production tools for industries, AI is rapidly evolving from an interactive chat tool into a universal productivity force capable of autonomous planning and task execution.
The Spotlight Shifts from Parameters to Agents
At the exhibition halls, major industry players prominently featured AI Agent products as their centerpiece exhibits. Attendees queued not to "chat with AI and compose poetry," but to "have AI complete specific tasks."
The Tencent booth, themed "Hey, my AI Buddy," showcased an AI Agent matrix that attracted a large audience. Staff demonstrated how its WorkBuddy agent could autonomously break down tasks, utilize tools, and deliver complete results based on simple voice commands like "organize a WAIC visit guide" or "build the framework for a Shanghai AI heatmap website," achieving full automation from instruction to finished product.
Booth staff noted that this workplace agent platform, launched in March, now boasts the highest daily active user count in China for office efficiency agents. Nearby, the operating system-level personal AI assistant Marvis was also highly popular, allowing users to remotely instruct their office computers via phone to perform tasks like system cleanup, file organization, and meeting mode switching.
Long-time exhibitor Transwarp Technology focused on the "data foundation" for agents. A preview version of its GPU-native cognitive database drew a crowd of professional visitors. Live tests showed that this database achieved orders-of-magnitude performance improvements over mainstream CPU-architecture databases when handling large-scale data analysis queries. Designed for AI Agents, this next-generation database integrates data analysis, knowledge retrieval, and long-term memory management on a GPU-native base, aiming to solve latency bottlenecks in agents' high-frequency data access.
From Office to Industry: AI Penetrates the Entire Production Chain
The transition of AI from "talking" to "doing" is not confined to demos; it is being validated in numerous real-world applications.
In the media sector, WorkBuddy has joined this year's WAIC reporting team for China Media Group, becoming its first "AI journalist assistant." It can generate a professional interview outline within 30 seconds, assisting with topic planning, data organization, and draft writing, supporting journalists in intensive conference coverage and pioneering a new "one journalist, one AI assistant" production model.
In enterprise collaboration, agents are deeply integrated into native workflows. The enterprise WeChat Agent "Dayuan," debuted at the conference, is embedded directly into the workflow. Users can swipe left to activate it for automatically summarizing key discussion points in group chats, suggesting replies, extracting data conclusions, and helping businesses document service records, track follow-ups, and derive business insights to enhance customer service and internal collaboration efficiency.
In office and creative fields, Sensetime's agent products have achieved large-scale validation. Its Xiao Huan Xiong office agent, enhanced with multimodal long-term task capabilities, has seen rapid user growth, now serving over 20 million users with 3 million weekly active users, more than 8,000 enterprise clients, and empowering 650,000 "super individuals." In content creation, its Seko multimodal creation agent has attracted over 1 million creators and 1,500 enterprises within a year of launch, with paid users averaging over an hour of daily creation time. Tasks once requiring professional teams can now be systematically delivered by ordinary creators using agents.
For entrepreneurs, lightweight AI tools continue to lower the productivity barrier. Hehe Information's AI Tools collection packages capabilities like document recognition, business information queries, and multilingual processing into tools for various startup scenarios, including foreign trade, finance/legal, and tech entrepreneurship, enabling small teams to perform professional-level information processing for tasks like data handling, customer discovery, and preliminary risk assessment at low cost.
Value Anchored in Tasks: AI Enters the Productivity Realization Phase
During the conference, numerous industry executives and experts highlighted that the AI industry is transitioning from a technology parameter race to a new stage of realizing tangible value.
Li Kan, Vice President of Tencent Group and General Manager of its East China Headquarters, stated, "Currently, AI is evolving from a tool to a partner, from 'generating content' to 'completing tasks,' from single-point capabilities to a systemic ecosystem. Tencent remains committed to practical iteration and efficiency priority, working with partners to explore the vast potential of digital applications in the era of AI Agents." He added that the goal is to make AI a practical productivity tool accessible to everyone, deployable by every organization, and beneficial to every industry.
In his keynote speech, Xu Li, Chairman and CEO of Sensetime, proposed that AI development is continuously broadening capabilities once reserved for professionals to meet universal needs, and the industry's metrics for value are changing. "Previously, the focus was on how much computing power and how many tokens AI consumed. As AI's service boundary expands to super individuals, the core future billing metric should be the price of task completion." In his view, the shift from measuring Tokens to Tasks represents a new industrial economic logic.
In his keynote, Yin Qi, Chairman of StepFun and Qianli Technology, observed that by 2026, model capabilities are crossing a critical threshold, with AI advancing from executing tasks for mere seconds to working independently for dozens of hours, positioning the industry at the foothills of Artificial General Intelligence. He proposed that agents are becoming the smallest unit of productivity, where their capability equals the product of model capability and agent operating system capability. "The greatest significance of the Agent era is not replacing humans, but amplifying them," Yin Qi stated, suggesting that agents will nurture a new generation of AI-native builders and organizations, ultimately amplifying everyone's potential tenfold.
From parameter competition to task delivery, from casual chat to production execution, this year's WAIC clearly outlines the evolution of the AI industry. As AI ceases to be merely a conversationalist on a screen and becomes an assistant on the office desk, an engineer in the factory, and a partner for entrepreneurs, the era of AI as a universal productivity tool for all is accelerating its arrival.
Comments