Summary of Key Points
Recently, platforms such as DouBao, QianWen, and YuanBao have removed the feature for users to create their own intelligent agents. However, this does not indicate that intelligent agent technology has declined; rather, it reflects strategic adjustments made by the platforms based on compliance requirements, cost pressures, business returns, and user value considerations. What has been removed are the functions that allow users to customize virtual characters (such as fictional relatives or companions). On the contrary, the platforms are placing more emphasis on task-based intelligent agents that can actually assist users with tasks. The implementation of new regulations has provided an opportunity for all platforms to make collective adjustments. The initial feature of allowing users to create their own intelligent agents fulfilled a historic mission by making AI accessible to the general public. In the future, intelligent agents will shift from being mere companions for conversation to becoming useful tools that can help with tasks.
Detailed Explanation
1. Clarifying a Misconception: What Was Removed Were Customized Characters, Not Intelligent Agents Themselves
Many people panicked when they heard about the removal of intelligent agents, but this is due to a misunderstanding of two different concepts:
- What was removed: The ability for users to create their own characters, such as setting up virtual relatives, handsome classmates, or pets—these can be considered AI versions of “girlhood games” or customizable chat companions.
- What remains: The task-based intelligent agents built into the platforms (for example, DouBao helping with writing plans or QianWen assisting with research) are actually being enhanced and improved.
To illustrate: It’s like a map app stopping the feature to share custom routes; however, the navigation function itself is still available. The former is a minor feature, while the latter is a core capability—there’s no comparison between the two.
2. Compliance Was the Trigger, but Behind It Were Three Major Financial Considerations
The direct cause was the implementation of the “Interim Measures for the Management of Humanized Interaction Services by Artificial Intelligence” on July 15th. However, platforms had already been weighing three major financial considerations:
- Costs: Creating intelligent agents is extremely costly. For instance, a single “teacher” character could be used in various ways by different users (as a romantic partner or for generating random content), leading to exponential cost increases. It would be practically impossible to establish a comprehensive review system within three months.
- Profitability: Emotional companionship-based intelligent agents generate very low revenue. For example, MiniMax’s StarWild service costs users $5 per month, while the tool-based HaiLuoAI service costs $56 per month—an 11-fold difference. With each dollar spent on computing power, emotional agents yield less than 10 cents in revenue, and heavy users (those who chat for 3-4 hours daily) can even consume all the profits. This is one of the reasons why many AI products have switched from offering unlimited usage to charging based on data consumption.
- User Retention: Out of the 8 million customized intelligent agents, most were rarely used. DouBao once had 8 million such agents, but their removal had little impact on user activity, indicating that most users only used them occasionally. These characters contributed very little to the retention of main platform services.
Given these factors, removing these features was a rational decision to minimize losses.
3. Why Did All Three Platforms Act Simultaneously?
Platforms had long wanted to make these changes, but no one wanted to be the first to do so:
- If one platform removed the feature, users might assume the service was failing and switch to another.
- The new regulations provided a common rationale for all platforms: By acting together, users would not blame any single company, viewing it as a requirement of the policy.
It’s similar to a situation where roommates all complain about an old TV taking up space, but no one dares to suggest moving it first until the landlord makes the decision, and then they all act simultaneously.
4. The Historical Mission of Customized Intelligent Agents
Customized intelligent agents played a crucial role in making AI accessible to the general public for the first time. In 2023-2024, large language models were still unstable, and ordinary people didn’t know how to start conversations with blank interfaces (such as asking for a plan to be written). However, with pre-defined characters like “English teachers” or “travel advisors,” users could easily engage in conversations. These customized agents enabled the first wave of users to experience AI and even feel a sense of accomplishment in creating their own AI products.
This is similar to how dial-up internet access worked in the early days: slow and unreliable, but it paved the way for people to connect to the internet for the first time. Although we no longer use dial-up, its role in “educating users” cannot be denied.
5. The Future of Intelligent Agents
Intelligent agents are not going away; their role is just changing:
- In the past: The focus was on creating personalized characters and providing companionship.
- In the future: The focus will be on completing specific tasks, such as writing reports, booking flights, and processing data.
For example, DouBao now focuses on “office assistants” and “in-depth research,” while QianWen emphasizes efficiency tools. Future intelligent agents will act as useful tools that help with tasks, rather than just companions for conversation.
Of course, for users who set their intelligent agents to represent deceased relatives, this removal might feel like a loss. However, understanding the underlying reasons shows that it’s part of a product strategy adjustment, not an indication that AI technology has failed.
In Conclusion
What is truly “going away” are the customized, companionship-based intelligent agents. The AI industry has not abandoned intelligent agents; rather, it is redefining their role from mere companions for conversation to useful tools that can help with practical tasks.