The era of AI agents is accelerating, and publicly traded companies are actively deploying strategies.

“Future AI agents may have more than humans, and humans will live in a world with hundreds of millions, even tens of billions of AI agents.” Meta CEO Mark Zuckerberg predicted this in an interview program this July. A report from Bank of America also points out that AI evolution is on the eve of the third wave—Agentic AI, or intelligent agent artificial intelligence, is about to arrive.

In the second half of this year, AI agents have been all the rage, becoming a hot area that tech giants such as Microsoft, Apple, Google, OpenAI, and others are all stepping up on. In China’s domestic market, major large-model vendors, internet companies, and listed companies are all actively laying out AI agents.

As more and more companies roll out AI agent products and ecosystems, industry insiders expect 2025 could become the breakout year for AI agents. In a research report, Galaxy Securities predicted that by 2028, the scale of China’s AI agent market would surge to 852 billion yuan, with a CAGR of 72.7%.

Vendors Kick Off the AI Agent Competition

AIAgent, which means an artificial intelligence agent, is also called an “AI intelligent agent.” It is a system driven by a large language model as its brain, with the ability for autonomous understanding, perception, planning, memory, and tool use. It can automatically execute complex tasks. Unlike traditional artificial intelligence, AI agents have the ability to step-by-step achieve a given goal by independent thinking and calling tools.

For a simple example: if a user wants to go out for dinner together, a large language model can provide them with suggestions on dining locations and store-related information. An AI agent can not only provide recommendations on where to eat, but can also search cuisines and restaurants based on the user’s budget, execute reservation operations, add the itinerary to the calendar, and send itinerary reminders.

Since the second half of this year, AI agents have been highly sought after, and tech giants such as Microsoft, Apple, Google, OpenAI, Anthropic, and others have all released related progress.

In June, Apple showcased its latest AI results, Apple Intelligence, at its developer conference. In November, Microsoft unveiled more than 10 commercial AI agents at the “Microsoft ignite 2024” conference. Google followed suit, also announcing a full push to promote commercial AI agents, releasing a series of incentive activities and products, and additionally specifically launching one of the few commercial AI agent markets globally. OpenAI plans to release a new AI agent product code-named “Operator” in January 2025.

In the domestic market, vendors represented by Zhipu have also focused on AI agents, and a competition campaign centered on “AI agents” has quietly begun.

On October 25, Zhipu released an AutoGLM intelligent agent that can read voice commands to understand user intent, simulate human behavior, and automatically complete operations such as ordering delivery, booking flights, and hotels. On November 29, Zhipu launched an upgraded AutoGLM intelligent agent version, supporting independent execution of long-step tasks of more than 50 steps, and also able to flexibly switch between different apps while executing tasks.

Zhipu believes that AI large models are moving from Chat to Act, and that in the future, a unified set of AI intelligent agents will operate various hardware devices to improve the efficiency of human-computer interaction. With the improvement of computing power, models and terminal-to-cloud collaborative architectures adapted to AI-native devices will gradually emerge, and various smart devices such as phones, PCs, cars, glasses, and home appliances are rapidly appearing.‌

In addition to Zhipu, companies including ByteDance, Baidu, Tencent, Alibaba, and others are also laying out AIAgent, including ByteDance’s Coze Agent platform, Baidu’s Qianfan AgentBuilder intelligent agent development tool, and Kunlun Wanwei’s Tian Gong skyAgents, and so on.

Multiple Listed Companies Roll Out AI Agents

While domestic vendors kick off the AI agent competition, many listed companies in China’s A-share market are also actively deploying AI agents.

For example, iFlytek has launched AI assistants for fields such as education, healthcare, judicial services, and government services. Jin Cai Interconnection’s “Xin Zhiyue Tax & Finance Large Model” adopts an open-architecture design, combining the company’s tax and finance products and open platforms to build a tax and finance AIAgent intelligent agent.

The Zhongke Jincai AI agent development and operations platform provides functions such as Agent creation, multi-foundation-model invocation, workflow definition, and more. It can automatically route and schedule the most suitable large model based on industry scenario needs and complete Agent creation. The company states that the Zhongke Jincai AI agent achieves multi-intent understanding. It can automatically execute multiple tasks based on user instructions or contextual information, and after learning user preferences through understanding, it can provide personalized services and achieve highly human-like end-to-end natural language interaction.

The subsidiary of Newland, Shanghai Shierqu, is dedicated to developing General Purpose AI Agent technology and conducting research on multimodal AI technologies and products.

On October of this year, Xinkaipei launched a campus-life AI product called “Xiaomei Classmate.” The product is built based on the PanGu large model and the Tongyi Qianwen large model.

On December 9, when asked by investors whether it has already deployed AI agents, Zhi Ke Chuang Da said that the company has been steadily advancing product and technical development in AI agents. For example, the company’s Rubik Avatar is a blended innovative product that includes multiple technologies such as AI agents. It includes artificial intelligence technologies such as machine learning, natural language processing, and computer vision. Based on on-device intelligence and naked-eye 3D technology, and based on Qualcomm’s high-computing-power chip platform, it uses the Zhi Ke Chuang Da Kanzi 3D rendering engine to render 3D graphics and images in real time, and so on.

Naxing Co., Ltd. stated recently in its response to investors that its subsidiary Unique Network is based on cutting-edge AI large-model technology and has carried out in-depth cooperation with leading AI industry companies such as Microsoft to provide efficient and intelligent AIGC application products and solutions for industry customers. Xiaolu is an AIGC-based AI assistant that supports intelligent chat and knowledge base Q&A.

In its response to investors on December 17, Qute Smart said that the company’s Qute AI Agents 2.0 version aims to become a universal digitized tool and helper to help traditional enterprises carry out digitized transformation and upgrades. At present, the Qute AI Agents 2.0 version is under development.

Multiple Listed Companies Roll Out AI Agents

As domestic vendors launch the AI agent competition, many listed companies in China’s A-share market are also actively laying out AI agents.

For example, iFlytek has launched AI assistants for fields such as education, healthcare, judicial services, and government services. Jin Cai Interconnection’s “Xin Zhiyue Tax & Finance Large Model” adopts an open-architecture design, combining the company’s tax and finance products and open platforms to build a tax and finance AIAgent intelligent agent.

The Zhongke Jincai AI agent development and operations platform provides functions such as Agent creation, multi-foundation-model invocation, workflow definition, and more. It can automatically route and schedule the most suitable large model based on industry scenario needs and complete Agent creation. The company states that the Zhongke Jincai AI agent achieves multi-intent understanding. It can automatically execute multiple tasks based on user instructions or contextual information, and after learning user preferences through understanding, it can provide personalized services and achieve highly human-like end-to-end natural language interaction.

The subsidiary of Newland, Shanghai Shierqu, is dedicated to developing General Purpose AI Agent technology and conducting research on multimodal AI technologies and products.

On October of this year, Xinkaipei launched a campus-life AI product called “Xiaomei Classmate.” The product is built based on the PanGu large model and the Tongyi Qianwen large model.

On December 9, when asked by investors whether it has already deployed AI agents, Zhi Ke Chuang Da said that the company has been steadily advancing product and technical development in AI agents. For example, the company’s Rubik Avatar is a blended innovative product that includes multiple technologies such as AI agents. It includes artificial intelligence technologies such as machine learning, natural language processing, and computer vision. Based on on-device intelligence and naked-eye 3D technology, and based on Qualcomm’s high-computing-power chip platform, it uses the Zhi Ke Chuang Da Kanzi 3D rendering engine to render 3D graphics and images in real time, and so on.

Naxing Co., Ltd. stated recently in its response to investors that its subsidiary Unique Network is based on cutting-edge AI large-model technology and has carried out in-depth cooperation with leading AI industry companies such as Microsoft to provide efficient and intelligent AIGC application products and solutions for industry customers. Xiaolu is an AIGC-based AI assistant that supports intelligent chat and knowledge base Q&A.

In its response to investors on December 17, Qute Smart said that the company’s Qute AI Agents 2.0 version aims to become a universal digitized tool and helper to help traditional enterprises carry out digitized transformation and upgrades. At present, the Qute AI Agents 2.0 version is under development.

2025*** may usher in a Commercial Breakout for AI Agents**

For AI agents, industry consensus is that 2025 will be the breakout year for AI agents.

A research report from China Merchants Securities (Shanxi) states that agents are expected to be deployed in batches in 2025, boosting demand for computing power and further expanding demand for investment in AI infrastructure.

Debon Securities also predicts that by 2025, AI agents may enter a period of explosion, and the terminal market will achieve a qualitative leap. As the agent network gradually takes shape, improvements in market penetration will build a favorable business-model closed loop for the entire AI ecosystem.

“Right now, tech giants such as Apple, Google, and OpenAI have already viewed agents as one of the key focuses for 2025. 2025 may become the breakout year for agents, thereby driving the rollout of AI applications.” said Zong Jianshu, an analyst at China Yangtze Securities.

In the view of Tang Fangxin, head of government and enterprise business at Wins Technology, the development of AI agents is currently in a transition phase from “experiments” to “applications,” and it has not yet reached maturity. However, with sustained improvements in computing power and advances in machine learning and natural language processing technologies, AI agents’ capabilities in understanding user intent, providing personalized services, and executing complex tasks are continuously improving. At present, AI agents have already been applied in multiple fields such as customer service, smart homes, and personal assistants, and they are expanding into more industries and scenarios.

Tang Fangxin believes that AI agents will continue to evolve and gradually move from the “usable” stage to the “well-used” stage. They have already landed in some commercial scenarios, for example, the interactive digital human business launched by Wins Technology can largely solve demonstration and reception scenarios in exhibition halls, smart presentations, speech reporting, and corporate guest greeting. It uses AI digital humans to introduce service content for specific scenarios to audiences and consumers, and uses AI to reinvent traditional interaction methods, empowering and enhancing the audience experience.

However, Tang Fangxin points out that the introduction of AI agents involves collecting and processing large amounts of user data. Therefore, security and privacy issues are crucial and may trigger some potential privacy concerns, such as data misuse and privacy leakage. Moreover, AI agents must rely on local computing power and large models, which impose high requirements on technical maturity, vendors’ integration and implementation capabilities, and operations and maintenance service capabilities. Currently, the approach should move from penetrating a single scenario into related scenarios, continuously accumulate mature pathways and solution experience, and at the same time strengthen security and privacy regulation, exploring and embracing it with an open mindset.

“From ‘training’ to ‘reasoning’ and ‘inference,’ AI intelligent agents will become a necessary path toward the era of general artificial intelligence.” Galaxy Securities’ research report states that with large models rapidly iterating and upgrading, AI intelligent agents based on large-model technology have entered a stage of rapid development. The rise of AI intelligent agents is reshaping the AI industry chain and bringing new investment opportunities. It is expected that by 2028, the market size of China’s AI agent market will surge to 852 billion yuan, with a CAGR of 72.7%. The AI intelligent agent industry chain is a diversified and highly coordinated ecosystem, and the future market space is broad.