Based on the author’s speech at the 2025 Digital Water Innovation Forum, this article delves into the development trend of the AI Agent industry and its far-reaching impact on the large water industry. The article elaborates on the definition, classification, and industrial ecology of AI Agent, analyzes its specific application scenarios and challenges in the large water industry, and presents readers with broad prospects for the application of AI Agent in the water field.
The following is a speech I gave at the 2025 Digital Water Innovation Forum, compiled with AI based on the speech PPT. Special thanks to Dr. Guo Yufeng, Deputy General Manager of Three Gorges Smart Water Technology Co., Ltd., for his invitation.
Hello everyone! Special thanks to the invitation to the conference, I was able to chat with industry experts and partners about the hot topic of AI Agent! I worked in the Water Resources Bureau for two years after graduating from my undergraduate degree, but then mainly did industrial research and investment. Not long ago, Mr. Yufeng told me that now AI bosses are eyeing this field, so today I hope I will talk to you about AI Agent from the perspective of industrial development, and more importantly, see what changes it will bring to our big water industry!
What exactly is AI Agent? For example, it’s like suddenly having a “super intelligent assistant” around you – using a large language model as the “brain”, which can “understand” needs and plan actions, and also has its own “memory” function, which can call various tools to do the job beautifully. In the future, it can take the initiative to help you solve the problem without you repeatedly telling you, isn’t it cool to think about it?
Looking at the classification of AI Agents, there are many doorways. According to the scope of tasks, there are “all-round players” who can help with anything, and there are also “industry experts” who specialize in a certain field; According to the service objects, there are 2B models for service enterprises and 2C models for individuals; According to the content of the task, from chatting to relieving boredom to sales customer service, there are simply a variety of things, only you can’t think of it, you can’t do it without it!
When it comes to the industrial ecology of AI Agent, it is like a huge river and lake! Upstream are “hardcore players”, chip giants Nvidia, AMD, Huawei, large language model developers OpenAI and Deepseek, as well as cloud service giants Amazon, Alibaba, and Microsoft, who hold core technologies, like “weapon forgers” in the rivers and lakes, providing key “equipment” for the entire industry. The midstream is a “technical school”, focusing on building platforms, developing frameworks, ensuring security, and silently escorting the birth and growth of agents. The downstream is the most lively, with various intelligent applications blooming, and the “big market” of the intelligent market, which is convenient for everyone to choose their own exclusive “smart assistants” on demand.
Over the years, AI has been seen more as a technology, and few people see it as an industry. But AI Agent is different after the emergence, it allows AI to truly enter all walks of life and solve practical problems, which is the real starting point of AI industrialization! Therefore, I often say that the AI Agent industry is a “key leap” for AI technology to move from the laboratory to the market.
Next, let’s talk about six major development trends in the AI Agent industry, each of which is closely related to our work and life!
The first trend, the explosion of demand! The 2B and 2C markets are in full bloom. For enterprises, using agents to reduce costs and increase efficiency and improve customer experience is simply the “secret weapon” to enhance competitiveness; For individual users, the future personal assistant agent market has huge potential and a proper 100 billion-level track. I boldly predict that within 5 years, the number of AI agents in the world may exceed 10 billion! In the future, there may be fewer and fewer apps on your phone, but your “smart assistant” will become more and more.
The second trend is that the main battlefield of competition has shifted from upstream to downstream. With the rise of open source models such as Deepseek, upstream competition has become fierce, and giants will inevitably set their sights on the broader downstream application market. In the future, whoever seizes the high ground first will be able to stand out in the three major fields of personal general assistants, personal or small B professional agents, and enterprise agentic solutions.
The third trend, AI Agent, will reshape many industries, especially HR-intensive ones. In labor-intensive work, such as customer service and data processing, agents can directly “take over” and complete tasks efficiently; In knowledge-intensive fields, such as wealth advisors and legal advisors, agents can become the “strongest assistant” for professionals, providing massive reference information and making work more efficient and accurate. I will focus on which sub-industries of the large water industry belong to the human resource intensive sub-industries, and these sub-industries are precisely the focus of AI Agent applications.
The fourth trend is that the corporate strategy revolves around “open source” + “interconnection” + “integration”. Take ByteDance and Google as examples, Byte has built a huge AI ecosystem through rich model layer and application layer layout; Google continues to launch innovative AI products with a powerful cloud platform and development environment. Open source can bring together the wisdom of global developers, interconnection makes agent collaboration smoother, and integration can maximize efficiency and profits in the early stages of industry development.
The fifth trend is that complex tool calls and multi-agent collaboration are becoming more and more important, which requires a unified “communication standard”. Just as people from different countries need a common language to communicate, Google’s open-source A2A protocol is the “common language” of the agent world, allowing them to work together seamlessly and exert greater value.
The last trend, government regulation must keep up. With the widespread application of AI agents, issues such as intellectual property infringement, data privacy protection, and liability definition will gradually be exposed. The European Union has introduced the AI Act to classify and supervise AI systems with different risk levels, and in the future, global regulations will be strengthened to ensure the healthy and safe development of AI agents.
After talking about the big trends, let’s focus on the impact of AI Agent on the big water industry!
Before that, let’s discuss a key question: Are AI Agents really “grabbing jobs” with us humans? In fact, it is not! Humans are essentially carriers of tasks, and agents are here to assist us in completing tasks more efficiently. What can really be replaced are those positions that only repeat a single task; But if there are tasks at work that cannot be replaced by agents, human beings still have value.
This requires us to think: which tasks are suitable for humans and which are suitable for AI agents or automation devices? Here I divide tasks into four categories from the two dimensions of “task complexity” and “fault tolerance”:
- High complexity – high fault tolerance For example, the overall planning of urban water systems involves the integration of multi-domain knowledge and dynamic decision-making, and the system has a certain degree of flexibility, which is suitable for humans and agents to complete such tasks collaboratively.
- High complexity – low fault tolerance The precise ratio of chemicals in the water treatment process requires extremely high precision, and any deviation may affect the water quality, requiring the integration of multidisciplinary knowledge, usually requiring the cooperation of humans, agents and automation equipment.
- Low complexity – High fault tolerance types such as customer service, copywriting, administration, etc., with fixed processes and tolerance for certain errors, are most suitable for agents to complete independently.
- Low complexity – Low fault tolerance types such as hydrology and water quality reports, simple to operate but demanding data accuracy, due to the “hallucination” problem of AI agents, are more suitable for handling with automated equipment or software.
Based on this classification framework, in our large water industry, labor-intensive, high-fault tolerance tasks with low per capita output are the easiest places for AI Agents to “show their skills”. Next, we look at the specific application potential from the perspectives of upstream and downstream of the industry and enterprise management scenarios.
The large water industry covers a wide range, from rainfall monitoring, water conservancy project construction, to water quality monitoring, sewage treatment, to water resources management, water conservation management, involving planning and design, construction and construction, operation services and other links, there are many upstream and downstream enterprises:
- Water conservancy and water engineering planning and design. For example, the Deep Water Planning Institute is responsible for industry engineering planning and design.
- Supply of core equipment and key materials. Enterprises such as Kaiquan Pump (water pump), Sany Heavy Industry (engineering vehicle), Weixing New Materials (water pipe), Shanchuan Wisdom (water meter), Ningshui Group (water meter) and Weipag (sensor and water management software).
- Construction and construction of water conservancy and water projects. For example, large engineering enterprises such as China Construction and China Power Construction.
- Water conservancy and water facility operation services. For example, BEWG, Shanghai Urban Investment, Chongqing Water, Yangtze River Power and other operating institutions. Supporting service institutions.
For example, industry regulatory departments, university research institutes, industry associations and financial institutions are of course also an important part of the large water industry.
Analyzing the data of typical listed companies in various sub-industries, it is found that the per capita operating income of upstream design companies (such as the Deep Water Planning Institute) and software and hardware system integration companies (such as Weipag) is about 60-700,000 yuan; about 1 million water operators; materials, equipment suppliers and construction enterprises are high, definitely more than 1 million.
What does a product manager need to do?
In the process of a product from scratch, it is not easy to do a good job in the role of product manager, in addition to the well-known writing requirements, writing requirements, writing requirements, there are many things to do. The product manager is not what you think, but will only ask you for trouble, make a request:
View details >
the per capita revenue of construction enterprises exceeded 3 million. Why is there such a difference? Because materials, equipment and construction enterprises have greatly increased per capita output through automated production lines and machinery and equipment, and design consulting and other fields still rely on manpower investment, so the labor efficiency is low. This is in the rapid development, the contradiction during the construction period of the project is not prominent, and today’s low human efficiency is likely to mean the survival of the enterprise.
Of course, from another perspective, this provides a broad space for AI Agent applications! It is emphasized here that scientific research and education are also human resource intensive, and from the perspective of fault tolerance, they are also a promising sub-industry for intelligent applications.
From the perspective of enterprise management scenarios, in production, sales, HR, CRM, knowledge management, financial management and R&D, the sales customer service, human resource management, administrative office and knowledge management of large water enterprises are particularly suitable for prioritizing the exploration of AI Agent applications, especially knowledge management, which is highly feasible and economical. Imagine how much efficiency can the agent quickly sort out and analyze massive water data and research results to provide instant support for decision-making?
However, our large water industry also faces challenges in applying AI Agent. The large water industry has the attribute of public utilities, the degree of marketization is relatively low, coupled with the fluctuation of local fiscal revenue, the application speed may not be too fast. While sub-industries such as planning and design, digital solutions, and scientific research are in urgent demand, insufficient payment capacity may be a real problem in the short term.
But don’t be discouraged! As the technology continues to mature, the penetration rate of AI Agent in the large water industry will definitely increase, and the efficiency of water resource management and allocation will continue to improve.
The emergence of Deepseek at the beginning of this year has completely ignited the enthusiasm for AI applications, and many water companies have set up water technology subsidiaries to develop vertical models of water affairs, and the “flywheel” of industrial supply-driven and demand-driven has turned. It is believed that in the near future, AI Agent will become a “powerful engine” for the upgrading of the large water industry, promoting digital water affairs to digital intelligent water!
Finally, I would like to remind everyone that this passion for AI is very valuable, but it must also be viewed rationally. The penetration of agents into the water industry will not be achieved overnight, and after the excitement, there are still many practical problems to be solved.
Digital and intelligent transformation requires not only enthusiasm, but also persistence and patience. End with a quote from Roy Amara: “We tend to overestimate the short-term impact of a technology and underestimate its long-term value. ”
