Personal Agent (hereinafter referred to as PA) has emerged as a new hotspot in AI. The intention economy it represents is now posing a challenge to the internet's attention economy.
Overseas, tech giants and startups are competing for the PA entry point. The market consensus is that the intention economy will replace the attention economy. In addition to Meta Muse and OpenAI Dots, a batch of startups including Instinct, Cue, Fambot, Poke, Skye, and Town have joined the battlefield.
In China, PAs are almost exclusively developed by internet giants and smartphone manufacturers, with startups largely absent. The underlying market consensus is that the intention economy is an extension of the attention economy; tech giants will continue to monopolize the intention economy through PA products, leaving startups with no access.
The differences between the Chinese and overseas markets can be analyzed through the PA service chain.
Figure 1: Personal Agent Service Chain
The first link is understanding the user. The PA needs to know who the user is, what they need, their surrounding social relationships, and a series of other information. Therefore, it requires users to authorize a range of personal life data, such as IM, social networks, personal notes, credit card bills, and so on.
The second link is initiating tasks. This may come from direct user delegation, such as ordering a cup of coffee or replying to work messages; it may also be proactively proposed by the PA, such as proactively ordering groceries for the user based on their recent health and budget conditions.
The third link is executing tasks, where the PA calls Agent tools to complete the task.
The fourth link is completing transactions, where the PA performs actions such as placing orders and making payments on e-commerce platforms.
After tasks and transactions are completed, the data returns to the first link of understanding the user, forming a closed loop. As the feedback data increases, PA products will gradually shift from passive services to proactive services.
In the above links, there is a huge gap between China and overseas markets in the three links of understanding users, initiating tasks, and completing transactions, which will also shape completely different competitive landscapes and ecosystems.
01. No Opportunities for Startups in China's PA Market, Rooted in User Data
In the first link of understanding users, the raw material is user data.
Overseas, users' life data is not within the tech giants' ecosystems. Bills, itineraries, calendars, and other data are scattered across emails, banks, and various vertical apps. Moreover, this data can be accessed as long as user authorization is obtained. Authorization equals access, which puts overseas startups on a similar starting line with tech giants in understanding users.
For example, Instinct's most core user data sources are emails, calendars, and credit cards. Its founder, Shinn, treats the time of users' first card submission, first password submission, and first sensitive document submission as the North Star metric. On this point, Meta and OpenAI also need to cross the same threshold of user trust. Recently, to improve the service quality of Dots, OpenAI has been focusing on obtaining authorizations for various associated user accounts to better understand and analyze users.
The situation in China is exactly the opposite; data is locked within the tech giants' ecosystems.
Tech giants are the winners of the attention economy, capturing a massive amount of user time. Within the internet giants' ecosystems, there is users' chat data, wallet payment data, and transaction order data. Smartphone manufacturers are the same; based on their mobile operating systems, they also possess a vast amount of users' contextual data.
However, startups cannot access the data within the tech giants' ecosystems. Even with user authorization, the accessible data fields are not comprehensive. This move hits the fatal weakness of PA startups.
02. No User Time, No Proactive Services
In the second link of initiating tasks, there are also significant differences between China and overseas markets.
The methods of initiating tasks are divided into passive user delegation and proactive PA delegation. Overseas, both coexist. Startups focus on user delegation, while tech giants focus on proactive services. Instinct's most typical scenarios are user-delegated flight and hotel bookings, with travel GMV accounting for over 50%. Muse's typical scenarios include proactive grocery shopping based on recipes, proactive reminders for itineraries and to-dos, and briefing pushes, all of which are proactive services.
In China, the market defaults to the belief that only proactive services have opportunities, while passive services lack room for imagination. Tencent's upcoming Handy Bot focuses on scenarios such as price-comparison shopping and itinerary monitoring, similar to Muse.
However, the foundation for PAs to provide proactive services is the user time represented by the attention economy. Sufficiently long user time not only accumulates user data but, more importantly, stimulates user needs.
In most scenarios, user needs are stimulated by the environment. For example, users may develop a need to purchase while browsing videos, or a need to travel while chatting with friends. Based on data from Instagram and WhatsApp, Muse can easily provide proactive services that meet user needs. Other internet giants and smartphone manufacturers are the same.
In contrast, startups lack user attention and can only provide passive services in the early stages. Only by accumulating sufficient user authorizations and historical service records can they potentially offer proactive services. Overseas PA startups are taking the passive service route, while China lacks market space for passive services, resulting in no opportunities for PA startups.
03. Separating Attention from the Transaction Closed Loop: The Only Chance to Kill the Attention Economy
The fourth link is completing transactions, where the biggest difference between China and overseas markets becomes apparent.
Overseas, the user time in the attention economy is concentrated in Meta and Google, while transactions are concentrated in top e-commerce platforms and independent merchant platforms, with both ends belonging to different companies. PAs that undertake user intentions must execute tasks across ecosystems, so they inevitably grow outside the ecosystems of all tech giants, thereby giving startups room to enter.
The structure in China is exactly the opposite; both attention and the transaction closed loop are in the hands of tech giants. Douyin's content stimulates the desire to buy, Douyin's e-commerce completes the order, and Douyin Pay completes the transaction; WeChat's social interactions build trust, and WeChat Pay and Mini Program e-commerce complete the payment. Therefore, PAs naturally grow within the tech giants' ecosystems, leaving no opportunities for startups.
Both in China and overseas, everyone believes that the core of traffic will shift from time to intention, and intention will become the most critical factor. Overseas, this represents that intention will replace time as the unit of traffic measurement, and the attention economy represented by advertising will perish. In China, this is a business transformation within the tech giants' ecosystems, where intention becomes the new core metric, and the transactions represented by intention replace advertising as the new monetization model.
The revenue of the intention economy can only come from transaction commissions; it is free for users, and merchants are charged a commission on completed transactions. The intention economy naturally conflicts with the business model of advertising. As Instinct's founder Shinn said, "Users are not commodities. If a Personal Agent influences users to buy things they don't want to buy, that would be a very dangerous world." Therefore, Instinct's product goal is to be free for life, firmly refusing to run ads, with all revenue coming from a unified commission charged to merchants.
Smartphone manufacturers have similar thoughts on this. Liu Zuohu, Senior Vice President of OPPO, believes that in the past, everyone thought about how to get exposure and how to invest in traffic; these advertising models will become things of the past.
Therefore, Chinese tech giants developing PAs is essentially a defensive move against the replacement of the advertising attention economy. Domestic PAs are first and foremost a defensive means, and only secondarily an offensive weapon.
04. The PA Strategies of Chinese Tech Giants
Following the release of Muse, domestic tech giants immediately followed suit.
Alibaba officially announced the launch of its Personal Agent product "Qwen Go!" yesterday, which is embedded in the Qwen APP, and the experience application page is now online; ByteDance is rushing to develop the Doubao Personal Assistant, with the product named Xiaodou; Tencent's product, internally codenamed Handy Bot, is being planned for launch as an independent app.
The PA strategies of tech giants are determined by the core assets they hold.
Alibaba is suited to build transaction closed-loop PAs: based on e-commerce, payments, and life services as the execution foundation, it directly guides user intentions to product selection, ordering, and service fulfillment, making the PA an intelligent entry point for its own transaction ecosystem.
ByteDance is suited to build content-stimulating PAs: relying on content and recommendation scenarios like Douyin, it discovers and stimulates needs from content consumption, then converts interests into searches, decisions, and transactions, allowing the PA to fulfill the demand realization within the content ecosystem.
Tencent is suited to build social service PAs: relying on WeChat's message touchpoints and Mini Program services, it connects user conversations, friend relationships, and service calls to complete the transaction closed loop within the WeChat ecosystem.
After tech giants exert their efforts, the space and window of opportunity for startups will further narrow. Chinese PA startups already have two fatal flaws: first, they cannot obtain user data; second, passive services in the cold start stage lack market space.
After the collective entry of tech giants, the survival space for startups is further squeezed. The positions left for startups in the PA track are more as providers of certain single-point capabilities.
For example, for PAs to complete the transaction closed loop, the construction of trusted infrastructure is also required. Trustworthiness is reflected on the one hand in the PA's ability to execute tasks properly, and on the other hand in whether its proactive service capabilities always fall within controllable, authorizable, and traceable boundaries. The latter requires the construction of various infrastructures such as payment protocols and standards, and security standards.