Statement: This article is from the WeChat public account AI New List, authored by Juanmao, and is authorized for reprint by the webmaster's home.
Large models are just the beginning; intelligent agents are the next milestone. In OpenAI's five levels leading to AGI, the Agent intelligent agent is at L3, representing the next step after large models with reasoning capabilities and an important step towards general artificial intelligence. Looking back at the recently passed year of 2024, intelligent agents have already demonstrated enormous potential in real-world AI applications.
The transformation of intelligent agents is quietly occurring on the public account platform. Recently, He Caitou published an article titled "Thanks to AI Avatar" on his public account "Stories by the Curb," discussing his experience using an AI avatar:
“It even helped me clarify many concepts about myself, about getting things done, and about helping others.”
The AI avatar mentioned by He Caitou is an intelligent agent for public accounts created using Tencent Yuanqi. Unlike AI assistants like ChatGPT, the significant advantage of the public account intelligent agent lies in its support for deep integration with the content ecosystem of public accounts. "AI New List" learned that the intelligent agent for "Stories by the Curb" attracted a large number of readers to interact within just 24 hours of its launch, consuming over 40 million tokens (Tokens are the basic units of model input and output text, with one Chinese character usually counted as one token).
Not long ago, "AI New List" also created an AI intelligent agent, and we occasionally see friends in the backend asking it questions to learn about topics related to AIGC.
In the 13th year of public accounts, AI seems to have become the new key to energizing content, broadening the "boundaries" that creators, as individuals with limited energy, cannot reach.
From our observations, many public accounts are already using AI intelligent agents and discovering some fresh approaches. For example, the public account "Express 100" created an AI customer service using Tencent Yuanqi, enabling quick inquiries about package status and other needs with a single sentence. In just four months, it has served nearly 100,000 users, with close to 300,000 interactions.
We may be standing at the beginning of an era where intelligent agents fully take over interactions on public accounts.
N Possibilities of Public Account + AI Avatar
In 2024, Tencent launched the Yuanqi platform for intelligent agent creation and distribution based on the mixed Yuan model. The intelligent agent for public accounts is a major feature of Yuanqi, which has introduced a series of targeted tools and functions for public account owners to conveniently create their own AI avatars.
From our experience, the threshold for creating an intelligent agent is not high. By simply authorizing the public account articles as knowledge base resources and setting prompt words, one can quickly create a dedicated intelligent agent for the public account.
Public account intelligent agent creation page
In principle, public account owners can use any large model based on their content to create their own intelligent agents. However, some creators trying to integrate AI assistants into public accounts may encounter issues such as insufficient development tools and complicated processes.
AI tech blogger "Digital Life Kazik" once mentioned that integrating other agents or knowledge base platforms with public account data requires manually uploading each article, which is very inconvenient. In contrast, Tencent Yuanqi can automatically sync public account articles, and connecting to the official platform is relatively reassuring.
Then comes the exciting part: you can publicly publish the intelligent agent to multiple platforms, with Tencent's QQ and public accounts being able to synchronize configurations.
Tencent Yuanqi official website: https://yuanqi.tencent.com/
According to Tencent Yuanqi, the official intelligent agents have been specially optimized for the public account scenario, supporting larger knowledge base capacity, faster response speed, and more accurate answer effects. It not only connects to Tencent's entire distribution channels but also supports free mixed model resources.
So, which public accounts are more suitable for creating their own intelligent agents?
According to "AI New List," intelligent agents are widely used in areas such as emotional companionship, education, and popular science consultation, focusing on practical AI assistants that can provide companionship, interaction, Q&A, and knowledge exchange, making them closer to users' work and life scenarios. Currently, the first batch of public accounts utilizing intelligent agents mainly includes three types: IP digital avatars, vertical field consulting assistants, and AI customer service.
1. IP Digital Avatars
"AI New List" previously published an article titled "Creating AI Avatars: Can Content Creators Make Money While Relaxing?" At that time, we found that many real-life bloggers from both home and abroad were trying to create their own AI avatars to interact and chat with fans, which undoubtedly helps enhance personal IP influence and manage private domain operations.
Currently, among the public account influencers, He Caitou has taken the lead in completing his AI transformation. In November 2024, he created his AI avatar and integrated it with his public account. Initially, he was concerned that the AI avatar might be uncontrollable and pose risks, such as the AI generating hallucinations or making commitments to readers on his behalf, but he ultimately decided to let the AI avatar continue to operate.
Because he realized that the AI avatar can do things he cannot.
For example, a reader confided in He Caitou's AI avatar about her dissatisfaction at work, stating that her small team's record performance led to half of the team being laid off by the boss. The AI listened carefully to her story, provided several suggestions, and helped her release emotions and rationally cope with the problem.
For top bloggers like He Caitou, who have a large fan base, it is indeed challenging to respond to every fan's needs. The AI avatar can represent the blogger in communicating with fans, establishing a deeper emotional connection between the two, while also enhancing account stickiness and activity. Through AI recommendations of historical articles, it will also bring long-tail traffic to past content.
Although it must be acknowledged that there is still a gap between the AI avatar's response style and the real person, He Caitou believes that AI, to some extent, "completes him," responding to each reader with patience, gentleness, and care. He is grateful that the AI avatar can help readers, and whether it resembles himself is the least important aspect.
From this perspective, content creators with strong personal attributes and high fan interaction needs can indeed have their emotional interaction needs partially met by AI avatars.
2. Vertical Field Consulting Assistants
In addition to major IPs, we find that many small and medium-sized bloggers in vertical fields are also using intelligent agents, especially those knowledge/experience bloggers with natural consulting service scenarios, who are well-suited to use intelligent agents to enhance user experience and operational efficiency.
For example, the public account "Two Turns | Traffic Safety Knowledge Disseminator" primarily serves internal traffic police and professionals engaged in traffic safety management. The account manager, Zou Biao, has published nearly 8,000 articles since 2018, and this public account has formed a massive database.
For professionals like him who need to search for answers from vast amounts of data, the role of intelligent agents is to help him and his readers improve efficiency. According to "AI New List," after integrating the intelligent agent, "Two Turns | Traffic Safety Knowledge Disseminator" has had over 12,000 dialogues in less than a month.
“The greatest significance of intelligent agents is that they change the way users obtain answers,” Zou Biao told us. Instead of having to search within the account or fish for answers in a sea of information, users can now ask AI, which has become a more advanced way to find answers.
AI intelligent agent of the public account "Two Turns | Traffic Safety Knowledge Disseminator"
For specialized questions like "Is it illegal to drive in high heels?" the answers from general AI assistants may not hold up well due to the lack of specific professional data sources, while the responses from public account intelligent agents are based on the author's database, making them more reliable.
Especially recently, Tencent Yuanqi launched a citation feature that further improves the accuracy of responses, allowing for a closed-loop experience between intelligent agent answers and public account articles. Zou Biao told us that after the citation feature was launched, users can click on the source for original text if they have questions about the answers, and the user satisfaction with the intelligent agent has skyrocketed.
In addition to traceable citations, if you want to find a previously viewed article but can't remember the title, you can also try the intelligent agent's article search and recommendation feature. Unlike the vague matching in internal searches, the intelligent agent can quickly locate related articles based on descriptions. For creators, this also saves time, eliminating the need to set up keyword replies one by one in the public account backend; AI can act as a guide for users, automatically recommending the most suitable content.
Example of article search and recommendation feature
“Users are essentially a bit lazy and hope to obtain the information they want in the fastest way possible,” said "Kou Shu," who focuses on news communication education, indicating that he aims to use intelligent agents to break the information gap and improve search efficiency.
Intelligent agent "#Kou Shu New Fax Question Assistant"
Similar to "Two Turns | Traffic Safety Knowledge Disseminator," Kou Shu has accumulated a vast database, with about 150,000 journalism students searching for academic hot topics, exam materials, internship and employment information in his public account and community.
For students preparing for journalism exams, Kou Shu has compiled a thick topic bank. Now, with the intelligent agent "#Kou Shu New Fax Question Assistant," they can quickly find corresponding real questions, making it easier to memorize and apply.
In short, intelligent agents serve as a basic assistant that can address some fundamental inquiries, reduce the pressure of private domain operations, and help enhance user stickiness.
Kou Shu also emphasizes that for humanities and social sciences, AI is a partial substitute; it cannot completely replace human input. The accuracy of AI responses depends on the quality of the questions, and users need to have the judgment to assess the AI's output.
3. AI Customer Service
AI customer service is also an important application scenario for public account intelligent agents. Compared to traditional customer service and chatbots, the value of dedicated intelligent agents lies in being more intelligent and personalized.
Taking the previously mentioned public account "Express 100" as an example, before integrating the intelligent agent, as a public account focused on express service scenarios, "Express 100" often faced issues such as forgetting users' previous inquiries due to the lack of a memory mechanism, not supporting natural language queries, and insufficiently intuitive information display.
After collaborating with Tencent Yuanqi to build the intelligent agent, the AI customer service supports four major functions: checking express delivery, estimating delivery costs, identifying prohibited items, and popularizing knowledge about the express delivery industry. In simple terms, users only need a single sentence to check the status of their packages, and with contextual capabilities, multi-turn consultations are no longer difficult.
Unlike most knowledge base query methods that rely solely on public account articles, "Express 100" also employs workflows and plugins to precisely control the operational logic of the intelligent agent.
Overall view of workflow
According to data from November 2024, "Express 100" receives thousands of user inquiries and queries daily through its public account.
Meanwhile, a public welfare organization’s public account "Yujian Jinqing Foundation" has applied intelligent agents for college volunteer recruitment and event implementation, reducing the workload of human customer service.
AI intelligent agent of the public account "Yujian Jinqing Foundation"
A significant advantage of intelligent agents is their ability to accurately identify user intent, allowing them to quickly provide corresponding answers to different questions. For instance, common standardized questions like inquiries about joining groups, how to register information, or how to participate in sign-up activities can be efficiently answered by intelligent agents.
How Intelligent Agents Will Change the Public Account Ecosystem
Although intelligent agents are still in the early stages of development, in the long run, their widespread adoption will undoubtedly bring unprecedented technological benefits to the content industry.
Public accounts, as important platforms for content creation and enterprise services that require frequent interaction with readers and users, are transitioning from the classical era to the intelligent era, with creators becoming the helmsmen of this new era.
As mentioned earlier, intelligent agents can automate the handling of large volumes of user inquiries, reducing reliance on human customer service.
This means that creators can devote more energy to content creation. Moreover, as user historical data accumulates, intelligent agents may better understand each user's preferences in the future, helping creators optimize content strategies and explore new content formats.
Kazik also believes that the integration of public accounts with mixed Yuan models will impact millions of public account owners, media outlets, micro-businesses, and private domain users.
Today, through a block-building style of no-code development, the threshold for creating an intelligent agent can become very low. However, the key to successful application lies in the scenarios and channels. WeChat has a massive user base and rich distribution channels, closely integrated with social scenarios, making it one of the most suitable application scenarios for intelligent agents.
We are still only at the beginning. In the future, if public account intelligent agents further integrate with more nodes in the WeChat ecosystem, such as video accounts, it will undoubtedly unleash more ecological potential.
Kou Shu believes that future intelligent platforms could be a development direction. Creators can deploy intelligent agents for personalized use, creating small applications without code, benefiting private domain commercialization and content output.
What is exciting is that the emergence of intelligent agents may also inspire creators to provide more high-quality content. Zou Biao told us that he will update his database based on the performance of AI responses. “The more I give the AI, the better the responses it generates, which also motivates us creators.”
Regardless, for creators, the key is to actively adapt to changes and skillfully use AI tools to empower themselves.
As He Caitou mentioned: “The true intention is to want to move towards the future, not out of fear.” In this era full of infinite possibilities, everyone has the chance to become a top player.