The first cup of coffee this autumn was ordered by the intelligent agent.
Starting from September, Alipay’s AI App Zhi Xiaobao and Zhipu’s intelligent AutoGLM can help users order a raw coconut latte with less sugar and no ice. Honor even launched a big move - letting the intelligent YOYO order 2,000 cups in one go.
When multimodality equips an agent with "eyes" and "ears", it begins to show promise in operating capabilities close to that of a human housekeeper - this generation of agents begins to learn to help humans "play with mobile phones", from daily shopping to friends AI can help users complete everything from making comments to travel planning.
As a result, is the mobile Internet ushering in a new revolution in intelligence?
In the era of mobile Internet, super apps form a closed loop of traffic by integrating services, but the emergence of intelligent agents is expected to redefine the connection between people and services.
People are beginning to worry about whether this change will redefine the new landscape of technology companies: With the arrival of intelligent agents, will apps die?
The answer is that apps that cannot be killed will evolve with the help of intelligent agents.
Today, Super App is far from being a software, but an entrance to a lifestyle.
For example, using Alipay is not only for payment, but also an entrance to life scenes such as financial management, travel, medical care, and tourism; using Meituan is not only for takeout, but also a gathering place for local life such as restaurants, supermarkets, and movies; using Douyin, it is not only Short videos are a business ecosystem carrying massive video content.
In the past, in the era of mobile payment, these super apps ended up being "muddling", laying out QR codes, building mini programs, and building a network of digital services through openness. In the AI era, they can also connect with thousands of offline merchants and institutions and help tens of millions of merchants and institutions upgrade from digital to intelligent.
Only when the intelligent agent is connected with real user needs can it truly come to fruition. Whoever can build the next intelligent agent ecosystem that fully meets user needs can become the king of entry in the AI era.
"Things that respond to natural language and can complete many different tasks based on knowledge of the user are called agents. Not only will agents change the way everyone interacts with computers, they will be the next platform."
Bill Gates’s definition of intelligence is also the future we imagine in the AI era.
However, in the first half of the year, major manufacturers gathered together to bet on the 1.0 stage of the smart agent platform, and their real money investment failed to quickly make a big splash in the traffic pool.
Overseas, Open AI's GPT Store was launched as early as January this year, and Ultraman had hoped that it would become the next "App Store"; domestically, major manufacturers such as Byte, Baidu, and Alibaba have also successively released intelligent platforms, pinning their hopes on Create "Super intelligence" (super intelligence).
However, in the 1.0 era, limited by the development of multi-modal capabilities, the agent at that time was more like a eloquent AI dialogue robot. Although it could provide users with knowledge, it could only stop at obtaining suggestions.
Therefore, in terms of user stickiness, most people still maintain an "early adopter" attitude towards intelligent agents. Even with the overwhelming traffic from major manufacturers, the growth of smart agents has been weak in terms of subsequent performance. On the platform, no Super smart agent has been born so far.
In the final analysis, it is a large number of fake demands created by AI capabilities that do not address the real pain points of users.
Compared with the 1.0 stage, Agent 2.0 focuses on specific scenarios and tries to meet the "real needs" of users.
Previously, the B-side applications of AI agents were mostly focused on code writing and auxiliary creation, while on the C-side, intelligent agents such as user-oriented companionship and psychological counseling were derived. As of July this year, according to QuestMobile statistics, copywriting, workplace work and emotional companionship have become common directions for the implementation of intelligent agents in mainstream AIGC products.
According to statistics from the AI product list, this year alone, the number of intelligent agents has increased by 179,000, which is 1.5 times faster than the growth rate of App Store applications.
Source: QuestMobile
In the second half of this year, agents have shown many changes in multi-task collaboration.
"Today's large model intelligence is constantly evolving from simple applications to complex applications, especially in the expansion of agents to o1 reasoning models, so that the system gradually evolves to be able to continuously interact with the outside." said Zhipu COO Zhang Fan.
Ordering takeout and booking air tickets in just one sentence has become a reality:
In September, Alipay launched its first service-oriented native app, Zhi Xiaobao. As an AI life steward, it can help users undertake "food, clothing, housing and transportation". They can complete daily tasks such as ordering food, swiping subway codes, and hailing a taxi with just instructions. It can also be intelligently sensed. Based on the time and space used by users, it intelligently recommends services such as news podcasts, express delivery inquiries, and travel strategies.
In October, Zhipu launched the intelligent AutoGLM, which can independently select multiple apps to operate and help users complete mobile phone interactions.
Subsequently, mobile phone manufacturers also followed suit. Honor's YOYO smart assistant and vivo's Phone Use can help users complete cross-application operations through one-sentence instructions.
In the past, users needed to find massive functions in complex interfaces, which was equivalent to increasing the user's cost of use. Now, just by expressing needs through voice or text, the agent can directly access the service and push the desired service directly to the user.
At this point, cutting into the urgent needs of daily life, the intelligent agent 2.0 has found a breakthrough direction - the "housekeeper" intelligent agent.
From ordering takeaways, adding to shopping carts, to canceling automatic app renewals, manufacturers are trying to integrate smart devices into our daily necessities, further simplifying the interaction between people and services, and liberating users from daily interactions with machines. For example, "Zhi Xiaobao" has always emphasized that "things can be done with just a word."
Although many "AI butler products" currently on the market can provide a relatively limited number of AI services and cannot perform more complex and personalized tasks, this evolutionary direction of human-computer interaction at least allows us to see that technology We are moving in a new direction - in addition to dialogue, we can also let AI "look at my eyes and act" to make life simpler.
In the mobile Internet era, traffic is life. The emergence of intelligent agents will also reshape the rules of traffic distribution.
In the 1.0 era, technology companies from overseas to domestic are trying to build super intelligent agent platforms to aggregate traffic through intelligent agents.
But the thinking of the 2.0 era has changed. Now, everyone is trying to turn the smart body into a "smart housekeeper" on the mobile phone and become a new entrance to connect users and services.
The most obvious manifestation of this change is the layout of mobile phone manufacturers. At the 2024 Consumer Electronics Show in Berlin, Germany, Fang Fei, President of Honor Product Line, said: "If the current smart assistant is manual driving on mobile phones, then the AI intelligent agent will be automatic driving on mobile phones in the future."
There may be predictions like this: when the smart agent on the mobile phone begins to learn to call required functions across applications, such as using Meituan to order takeout, opening Taobao to buy daily necessities, and by disassembling task scenarios, select different App operations to complete the task. Accordingly, the super App only needs to provide some interfaces for the intelligent agent to call. In the long run, the App will become a part of the intelligent agent's capabilities, and the traffic that should have flowed to the super App will also belong to the intelligent agent.
But in the diversified business era, competitive and cooperative relationships are the norm. On the one hand, mobile phones and super apps need to polish their AI products, use product competitiveness to win users, and compete for the initiative in new entrances; on the other hand, just as the prosperity of the mobile Internet is the result of everyone adding firewood and one day succeeding, the AI era The network of services is by no means monopolized by any one technology giant. Openness and cooperation are still the future of AI.
As Honor CEO Zhao Ming said, there is a collaborative relationship between the two. After finding the boundary point, everyone completes their assigned tasks through their own collaboration.
For mobile phones, if an intelligent agent wants to open up a complete service ecosystem, it requires the integrated supply of a large number of service resources.
As for apps, they can delve deeply into vertical scenarios, complete evolution with the help of intelligent agents, and renew many services in the AI era; at the same time, they can explore more ways to play with software and hardware linkage by cooperating with mobile phone manufacturers.
For example, as manufacturers rush to develop AI search products, community apps including Xiaohongshu and Zhihu are trying to create vertical search services through their long-term content advantages. Take Zhihu as an example. It has targeted the academic search track and launched a professional search function in Zhihu Direct Answer, becoming the first manufacturer to provide a one-stop solution for AI search and genuine paper library.
In the current craze of intelligent agents, ecological capabilities will also become the trump card and moat of apps.
With 4 million merchant mini-programs and more than 8,000 life service capabilities, Alipay's AI life manager "Zhi Xiaobao" can support calling taxis, ordering food, booking tickets, subway codes, checking express delivery, paying phone bills, and checking bills. and various life services - this kind of ecological integration capability is difficult to catch up with a pure intelligent platform.
At the same time, the current AI operations such as ordering coffee demonstrated by mobile phone manufacturers still use technical solutions based on screen recognition and simulation operations (you will see AI helping you view the screen and click buttons), which requires high performance of the mobile phone. , there are still problems such as slow speed and single service.
If you want AI to do better, you need supply changes on the service side - a large number of business organizations can also "AI", build their own intelligence, and then promote innovation in life services through open interfaces. Only when more merchants and institutions have intelligent agents can AI not stop at simple operations like ordering coffee, but can help you order more, order faster, order more accurately, and even help you find the most suitable one. of coupons.
Undoubtedly, just like millions of small programs were built in the mobile Internet era, building an intelligent agent ecosystem in the AI era is what WeChat, Alipay and other national apps are good at. Combined with the unique platform ecosystem, App can also become a new intelligent agent platform and break out of the AI melee with the help of differentiated services.
For example, after Tencent launched the smart assistant app "Yuanbao", it created the smart platform "Yuanqi"; Alipay also launched the smart development platform "Treasure Box", allowing merchants to use smart phones to provide users with more updated services. .
Take "Huang Xiaosong" as an example. It is an intelligent agent established by Huangshan Scenic Area on the Zhixiaobao platform. It can provide tourists visiting Huangshan with real-time tourist attraction guides, scenic hotel recommendations, power bank inquiries and other services.
In addition, App manufacturers can also jump out of mobile phones and interconnect with more smart hardware , such as AR glasses, smart speakers, smart cars, etc. In the future, AI will be everywhere, services will be available upon request, and the methods of human-computer interaction will be more diverse and innovative.
Previously, Doubao, a subsidiary of ByteDance, launched the Olla Friend, an AI smart headset, which provides users with an "AI friend" that combines functions such as portable know-it-all, English training, travel guide and emotional refueling station; it will be launched next week The new Rokid AR glasses released will also work with Zhi Xiaobao to launch functions such as AI taxi hailing, AI food ordering and voiceprint quick payment, covering more life scenes.
In the AI era, the reshuffle cycle will be further shortened. Zhu Xiaohu once said bluntly, "When everyone rushes into the hot spot, after 6 months, if you are not at the top, the hot spot basically has nothing to do with you." When the short-term hot spot comes, no one wants to leave the poker table first, AI The next generation of new mobile phones will be born, and Apps also hope to use AI to evolve again. The pioneers who are the first to deploy will undoubtedly win the next era.
But more importantly: those who travel alone are fast, and those who travel together are far away. In the AI era, no one company is dominant. There is competition, but there is even more cooperation. Openness and connection are originally the original meaning of the Internet. In the AI era, only when software and hardware are open to each other and countless intelligent agents are connected to each other can real changes be brought about.