Disclaimer: This article is from the WeChat official account "Silicon Star Pro," authored by Summer, and is republished on Zhanzhang Zhijia with authorization.
In 2024, contrary to original expectations, AI did not surge ahead.
In February, Sora emerged, sparking fierce competition in video generation models both domestically and internationally; by June, capital and entrepreneurs turned their attention to embodied intelligence, seeking to find a body for large models; in August, the emergence of OpenAI o1 opened up another path beyond the Scaling Law.
Reviewing the key milestones of AI in 2024 reveals that progress by top companies like OpenAI in foundational large models has slowed. Beneath the iceberg, the sparks of application have begun to ignite.
After technological breakthroughs come the democratization of technology. In 2024, the consensus across industries was almost unanimous: to bring AI into every household.
And this is no easier than developing a "ChatGPT."
Companies began to move: ByteDance's Bean Bun, Baidu's Wen Xiaoyan, and other large language model products, aiming to become universal conversational assistants, answering questions from users in all aspects; DingTalk, Feishu, WPS, etc., seized the office scenario, starting from the vulnerabilities of workers; Kuaishou's Kelin chose to start from the content entertainment scenario, finding a breakthrough point to revive old photos; Alipay, from the daily life service needs of ordinary people, quickly booking tickets, ordering meals, hailing rides, and querying nearby dining and entertainment through AI conversations, acting as a proxy for users...
Their commonality lies in finding a sharp scenario to enter, with the difference being the thin or thick integration of the scenario with AI.
Among the above scenarios, general conversation and entertainment scenarios are widely covered but not thick, with more scattered and personalized user needs, requiring gradual cultivation of user habits; office scenarios are thick enough, with very clear user needs, but the coverage is narrower, and higher requirements are needed for AI generation capabilities, requiring AI to delve deeper;
The life service scenario is a middle product of the two, not only having a wide enough coverage, almost everyone can use it with clear needs, but also not high requirements for AI generation effects, focusing more on whether the service itself is completed, seemingly achieving an accurate match between technology and scenario at this stage.
Taking Alipay's AI as an example, as a national-level app with 1 billion users, Alipay's life service scenario is naturally closer to ordinary people, also bringing more experimental fields for the popularization of AI. In this scenario, Alipay's AI product, Zhi Xiaobao, connects various service scenarios within Alipay, the key is not to generate beautiful articles or calculate complex formulas, but to call a taxi, order coffee, making AI capable of "execution" rather than just "Chat."
Alipay, which has chosen to root in the life service scenario, seems to be exploring a new path for AI popularization.
What is the best scenario for large-scale AI implementation?
Talking about AI implementation for so long, why can't it be realized? What kind of products can good technology make? Can good products solve user problems? Can solutions to user problems be used frequently? Facing the层层barriers between technology and products, products and users, users and scenarios, how can one open the任督二脉?
Some people choose to start from technology and push forward to scenario implementation, looking for nails with a hammer.
Apart from Kuaishou's Kelin bypassing the shortcomings of text-to-video technology and finding a scenario to bring old photos to life, successfully breaking through the circle, the popular MiaoYa APP in 2023 also found a scenario for text-to-image in portrait photos, but due to limited demand for these scenarios, the path to improving popularity and usage frequency is not yet obvious.
Others choose to start from the scenario and seek technical solutions.
Last year, DingTalk began to try to reconstruct itself with AI, from Magic Wand, AI search to 365 personal members, optimizing the daily workflow of workers, by reducing the complexity and bloat of work information and processes, thereby improving collaboration and production efficiency.
Similar to DingTalk, Alipay's AI is also an AI exploration based on its own ecological scenario.
As the largest life service super APP in China, in September this year, Alipay launched the AI life管家「支小宝」, connecting various service functions within Alipay. Over the past 20 years, Alipay has accumulated a wealth of service experience in multiple life areas such as payment, finance, transportation, and healthcare, among which, the transportation business has served over 8 billion trips, and the healthcare sector has served over 600 million people seeking medical treatment. As a result, Alipay has also become an "all-purpose treasure box" for ordinary people, which is the biggest foundation for Alipay's entry into AI.
Therefore, Alipay's「支小宝」does not focus its AI capabilities on writing and analysis, but on「practicality」, focusing on understanding user demands, calling up services, and achieving execution.
For example, for the commuting scenario of workers, users can use a sentence to call up the subway code or bus code, find nearby shared bikes; on the way to work, they can have AI buy a cup of coffee or milk tea according to their familiar taste, and pick it up when they arrive; when they are too lazy to type after work, they can just say a few words to call a taxi home, or search for nearby recommended restaurants, movie theaters, etc., as well as functions like topping up the phone bill, paying the electricity bill, and more.
The advantage of starting from existing scenarios is that users naturally have needs. Especially the life service scenario with wide coverage, providing a better environment for large-scale AI popularization.
Based on the launch of the Zhi Xiaobao product, Alipay also introduced a series of supporting measures to accelerate the evolution of Zhi Xiaobao, hammering AI deeper and more solid.
A key move is that Alipay has launched the "Intelligent Agent Ecosystem Open Plan" to the industry, and introduced the Intelligent Agent Development Platform "Treasure Box." Relying on professional intelligent agent construction capabilities, businesses and institutions can create exclusive intelligent agents in 0 code and as quickly as 1 minute, and publish them to Alipay Mini Programs, Alipay App, Zhi Xiaobao App.
The breadth of life service scenarios is one of its characteristics. This plan also means that Zhi Xiaobao is absorbing partners to join, further expanding its service scenarios.
Currently, Zhi Xiaobao has successively launched Huangshan Cultural Tourism Intelligent Agent "Huang Xiaosong" and Hangzhou Cultural Tourism Intelligent Agent "Hang Xiaoyi", starting with the cultural tourism segment scenario in the life service scenario.
"Hang Xiaoyi" is the first city-level cultural tourism intelligent agent in Zhi Xiaobao, based on the Alipay ecosystem, tourists can directly converse with Hang Xiaoyi, letting AI help with navigation, hotel reservations, scenic spot guides, and collecting discounts. From October 1st to 7th, "Hang Xiaoyi" served nearly 2.5 million tourists, with real-time scenic traffic queries and scenic route companion explanations being the most used services.
It is worth noting that "Hang Xiaoyi" has, in addition to functional services, leveraged the offline "tap + AI" method to lower the threshold for AI use. In over 40,000 service points across Hangzhou's A-level scenic spots, hotels, tourism consultation points, and major business districts, tapping the phone on the blue "Hang Xiaoyi Smart Stickers" can summon the "Hang Xiaoyi" intelligent agent on Zhi Xiaobao.
Judging from the data, the promotion speed of tapping within 2 months is 6 times that of face-swiping payment during the same period, serving over 50 cities including Shanghai, Chengdu, Wuhan, Changsha, Hangzhou, Qingdao, and jointly creating over 50 "tap + X" industry solutions with partners. Just as WeChat and Alipay's QR codes spread across China, bringing convenience to mobile payments. Today, through "tapping," AI has also built an open ecosystem that connects services with services, online with offline, bringing users "AI convenience."
Interconnectivity, a useful AI needs an open ecosystem
But it's not simple to make AI work for users, which is why, although life services are an important scenario, there are not many contestants in this track.
On the one hand, AI needs to be able to call up services, which means equipping AI with corresponding service capabilities. On the other hand, AI needs to understand users well enough to obtain coherent user behavior data preferences and provide accurate personalized planning, not doing one-point-at-a-time AI.
Both aspects rely heavily on the capabilities of the ecosystem; otherwise, AI is hard-pressed to cook without rice.
Just as Apple Intelligence is so eagerly anticipated by the public because Apple has integrated the AI ecosystem at the system level, giving developers more space to create AI applications, and through software and hardware synergy, integrating different scenarios that were originally separated within the system. For example, when a user takes a photo of a restaurant with the camera, the review page can show the restaurant's ratings; the user can ask Siri about the information in the photos taken by the camera and search online, and so on.
The power of the ecosystem was already demonstrated in the internet era's "mini-program" revolution. In WeChat and Alipay, the emergence of mini-programs not only expanded service capabilities but also connected users with scenarios, forming a tight and solid network, laying a solid moat for these two APPs.
And the intelligent agent is the "mini-program" of the AI era.
Similar to mini-programs, intelligent agents are the connectors between users and scenarios in the AI era. Under the network effect, the magnitude of scenarios is proportional to the magnitude of users. Only by opening the intelligent agent ecosystem can scenarios penetrate and services spread, attracting user aggregation and exerting greater effectiveness of AI.
The open ecosystem has been an important foundation for Alipay's development. Now, Alipay is continuing this route, expanding the ecosystem.
Currently, the "Intelligent Agent Ecosystem Open Plan" has made initial progress. In addition to cultural tourism intelligent agents, in the industries of government affairs and people's livelihoods, healthcare and health, Alipay has also gradually opened intelligent agent solutions.
Last November, Zhejiang Health Commission partnered with Alipay to launch the country's first digital health person「安诊儿」, relying on AI large model + digital human technology, providing users in Zhejiang Province with full-process accompanying services before, during, and after the consultation. So far,「安诊儿」has entered more than 1,000 hospitals in the province, serving over 10 million people, with a user satisfaction rate of 98%, as well as the official service platform of Jiangsu Province's government affairs, jointly creating the Xiao Su intelligent agent for the government affairs industry, and so on.
It can be imagined that in the future, every ordinary person will have an AI assistant with extensive knowledge, enough to understand you, and capable of helping you solve real problems, handling issues such as travel, medical consultations, insurance, financial management, auxiliary creation, etc.; more businesses and institutions can create exclusive intelligent agents with lower thresholds and costs, acting as AI customer service, AI accountants, AI marketing consultants...
Under the intelligent agent ecosystem, although user needs have not fundamentally changed, the specific services and the way they are provided have shifted from user search to AI understanding, judgment, and distribution. This not only optimizes user experience, making it easier for users, but also brings more imagination to businesses and institutions. After all, the more proactive AI is, the more it takes the initiative to explore user needs for incremental space. For example, for a single user need, AI can provide a series of service actions, completing user needs in a coherent manner.
For businesses and related institutions, this is a great opportunity to break down and reorganize their services with AI, opening up new user bases, and with over 4 million mini-programs, businesses need more such initiative.
Conclusion
Although Zhi Xiaobao has just been launched for a month and still has issues such as inaccurate service invocation, the good ecosystem soil and clear route make Alipay's AI full of imagination.
The mini-program carrier brought services from various industries from offline to online, allowing users to experience a comprehensive digital transformation. The advent of the AI intelligent agent era, from users finding services to AI intelligently connecting services, will inevitably bring new imagination to various industries.
For Alipay, continuous openness has brought experience and soil as it transitions from the mobile internet to AI intelligence. Alipay's rapid iteration and increased openness in AI have also found a scalable path for AI popularization.
No one can predict the future of AI, but some can create the future.
Execution is always the best solution, and perhaps it is also Alipay's answer in the uncertain context of AI development.