In 2003, Clayton Christensen put forward a point in "The Innovator's Answer" - successful innovation often has two footholds, one of which is to greatly lower the threshold of use, make it available to everyone, and greatly improve the efficiency ......
As predicted, over the next 20 years, this view was enshrined by major Internet pioneers and repeatedly tested in technological changes one after another.
Google and WeChat, the products of search in the PC Web era, now look up, the large models and AI that have been bubbling in the past year, in addition to bringing the highlights of the "life and death" 100-model war and the 100-million-level application battle, there is a new round of entrance war. On the water, there are already waves undercurrent.
01 A new round of entrance competition
Looking at the history of science and technology of the development of the Internet in the past few decades, it records dense business strategies and specific story cases, but if you look closely, all kinds of past are writing an industry truth - those who get the entrance will win the world.
What is an "entrance"?
This term is based on the PC Internet, which can be simply understood as the way for users to find information and solve problems on the Internet. From ancient times to the present, human beings have never stopped exploring the unknown, and correspondingly, people have never stopped chasing the entrance.
Whoever can become the entrance will be able to get a large number of users; And whoever can accurately aim for the future and grasp the needs of users will be the most likely to become the entrance.
Whether it is the traditional Internet era, the mobile Internet era, or the last wave of AI, "ambitious" Internet players regard the entrance as a battleground.
In the era of the traditional Internet, browsers and information portals are prevalent, and users mainly use search engines for information search, cache, index display, and guide to jump to major ** to actively obtain information. In the end, the monopoly of search engines became the entrance to the traditional Internet era, and the battles that were fought around it at that time, such as the game with Ali and the 3B war, seemed to happen yesterday.
In the era of mobile Internet, Apple has redefined the form of entrance with the help of "iOS+App Store", and human-computer interaction is no longer carried out through a unified search engine, but is replaced by one independent APP after another represented by WeChat.
These independent apps provide different content and services, if you want to get the corresponding information and services, you need to install the corresponding apps, so in the news, social, office, online shopping, daily life and other scenarios have their own "mountain" apps, if you want to complete a coherent operation of multiple scenes, the characteristics of the apps are not connected with each other is doomed to low efficiency.
Of course, various large manufacturers are also trying to expand the scale of imports in a platform-based way, but because the market pattern has been formed for a long time, it is difficult to break it for a while.
With the development of AI technology in recent years, the form of competition for entrances is also changing.
For example, Tesla, launched by Musk, which is never lacking in heat, can be regarded as a major entrance to intelligent driving and human-computer interaction. In addition, he bought Twitter for $44 billion in cash, ostensibly claiming to uphold freedom of speech, but under deep investigation, it was actually to seize the entrance.
What the next entrance will look like, no one knows. However, Geely Group and Weilai Automobile have invariably begun to make mobile phones, and Huawei and Xiaomi, the players on the mobile phone side, have also made efforts in the field ......of automobiles
If it is said that the big factories tried to achieve cross-scene comprehensive entrances through platformization, then the rise of car companies and mobile phone factories in the past two years has brought new ways to play the entrance war.
However, in the Internet era, the entrance battle may have different manifestations in different periods, but in essence, it is the competition for traffic.
Until, ChatGPT thundered, generative artificial intelligence entered the technology and even the public eye, and the underlying logic of the entrance changed.
02 Under the craze of large models, new rules, new wars
According to Leifeng.com, recently, IDC and DingTalk jointly released the "2024 Top Ten Trends in the Application Layer of AIGC"**By 2024, the development of the digital economy will give birth to more than 500 million new applications worldwide, which is equivalent to the total number of applications that have appeared in the past 40 years.
It can be seen that the AI model will bring changes of magnitude and new market logic.
Among the 500 million new applications, there may be a super application - providing users with one-stop service, which is also the key in the new round of entrance war.
In this regard, the pioneers have not tried before, but lack the corresponding technical capabilities, and this time, the disruptive changes that AIGC will bring to the form and format of application software are the most critical variables.
Everything is traceable.
At the beginning of last year, the explosion of ChatGPT ignited people's enthusiasm for AIGC, and all eyes were focused on the basic model, and the battle of 100 models was about to ......
By the end of the year, players were realizing that the field of base models for volume parameters could no longer be imagined.
A large model is like a car engine, and what people need is a car, so in addition to the engine, they also need all the necessary components such as the steering system, chassis, interior, etc. In order to give full play to the value and potential of large models, it is necessary to add a series of advanced technologies to the basic large models to give birth to AI native applications.
Large models have the ability to generate, remember, logic, and reason, which can provide a strong source of power for the research and development of next-generation applications. In addition, large models can also minimize the development threshold, simplify the development process, and improve development efficiency, even inexperienced developers can easily get started and customize an application according to their needs.
All of this provides the foundation for the upcoming 100-million-dollar application campaign. It is foreseeable that applied innovation will be the focus of the next war.
In foreign countries, Open AI, which entered the market first, and Google, which is long overdue, have begun to seize the entrance. Open AI's announcement of GPTS with great fanfare is like throwing a stone on the surface of the water, setting off waves that extend to the entire industry, and push the battle for entry to a climax.
On Open AI's GPTS platform, even a novice who doesn't know how to code and doesn't have basic computer knowledge can create an application that meets his own needs in 3 minutes through natural language interaction alone, making it possible to mass-produce applications.
In China, as early as the explosion of AI large models, manufacturers began to do various applications based on large models, such as office, medical, finance, education and other industries, new applications emerged one after another. Alibaba, and other self-developed large model manufacturers, are now increasing their investment in the application side.
For example, Robin Li has repeatedly emphasized that it is necessary to roll up AI native applications, so that the model can be valuable. Now, this battle is about to join a new player, DingTalk, which started early to carry out AI+ native reconstruction from the scene based on the Ali Tongyi Qianwen model, and now has made a set of strategic designs similar to GPT Store.
Of course, compared with self-developed large model manufacturers, DingTalk may not have many advantages in large model capabilities, but it has its own strengths, scenarios and data, and customers also have accumulated data and knowledge base; There are also clear customer needs.
These give DingTalk the opportunity to participate in the table in the battle for the entrance.
In fact, not only that, but the office scene that DingTalk is deeply cultivated also increases the possibility of DingTalk going further in the entrance competition.
03 Master the data, will give priority to the right to speak
Why does the office scene also have the opportunity to give birth to a super entrance?
The essence of the birth, development and progress of technology is to improve productivity and liberate human beings from labor, and there is a great demand in the office scene, and there is an urgent need to improve the efficiency of production and innovation.
According to IDC's research results, the top business benefits that enterprises want to realize through AIGC today include improved customer experience services, increased developer productivity, differentiated competitive advantage, and innovative business models.
Therefore, when entering the post-AI large model era, major manufacturers have begun to consider how to accelerate the implementation of large models and enter the stage of "volume" applications.
Focusing on these needs, I can't help but think that the ultimate form of large model landing may be AI agent (super assistant), like humans, "intelligent" software that can independently execute, make independent decisions, actively explore, self-iterate, and cooperate with each other, which can be regarded as the ideal intelligent assistant for humans.
This is indeed the case.
According to IDC's survey results, all enterprises believe that AI agent is the definitive direction of AIGC development. At the same time, 50% of enterprises have already piloted AI Agent in a certain job, and another 34% are developing plans for the application of AI Agent. For example, the GPT store planned by Open AI is aimed at creating an AI agent store.
Based on this, it can be predicted that AI agent is the mainstream form of large-scale model landing business scenarios.
The AI agent may be the first to be implemented in the office scenario.
On the one hand, knowledge management applications are the most popular application scenarios for AIGC organizations, and on the other hand, office scenarios are in urgent need of improving production and innovation efficiency.
This is also a point that is generally recognized by the industry.
At present, Microsoft, the world's largest office service provider, as early as the beginning of GPT4's "hanging", loaded its "family bucket" software with large model capabilities to generate Copliot functions, and raised the banner of a revolution in improving office efficiency around the world.
However, this is still an early form of large-scale model application, which is "+AI", which is what Ye Jun, president of DingTalk, said to add AI elements to the application.
Ye Jun, President of DingTalk.
Ye Jun believes that the development of generative AI and large model technology can be divided into three waves according to the development trend.
The first wave is the emergence of large models, represented by GPT models, which are the basis for the development of generative AI. The second wave is the innovation of the application layer, and the productivity scenario is the best bearer. For example, Microsoft Copilot and DingTalk released the AI magic wand in April last year, all of which make intelligence go from chat to work.
The third wave is the in-depth entry of AI into business scenarios, connecting with business digitalization and serving the real economy. This is the transformation of application "+AI" to "AI+", and the landing application of large models should be AI-native, which means refactoring, rather than simply "stitching".
At this point, DingTalk's ability is beyond doubt.
In April 2023, DingTalk will be connected to the Ali Tongyi Qianwen model, and the comprehensive intelligent strategy will be officially launched. By the end of the year, 700,000 companies were actually using it.
Those who enter first go fast.
Not only that, but yesterday, DingTalk deeply entered the business scenario with AI, implemented business digitalization into practice, and launched a new product of AI agent-AI assistant, which can be customized by everyone and every enterprise with personalized and exclusive super assistants.
DingTalk released an AI assistant.
According to Leifeng.com, DingTalk AI assistant has the ability to perceive, remember, plan and act, and also has the ability to perform tasks across applications, which can be seamlessly combined with rich third-party applications on DingTalk and self-built applications of enterprises.
Ye Jun, president of DingTalk, also said that an AI intelligent assistant platform and application market will be built, so that users, developers, and ISVs can develop personalized AI assistants with a low threshold on DingTalk, so that more excellent AI applications will emerge on DingTalk. "In the next three years, 10 million AI assistants will be created on DingTalk. "It's no longer a dream for anyone to be able to create an AI assistant.
04 Take the data and scene to break through, is there a play?
There is no doubt that AI agent will be the best AI application portal, which has become the consensus in the industry.
So, among many players, will DingTalk be the first to succeed and become a new entrance in this round of technological revolution? After all, the advantages of DingTalk are so obvious.
First of all, DingTalk has rich application scenarios and business data.
In the AIGC era, the key to the entry competition lies in the data, and for large models, the more high-quality data is fed, the more fully the performance will be optimized.
DingTalk is the first national-level work application in China that fully opens up AI, co-created with 700,000 enterprises, so it has rich resources in business data. In addition, these more than 700,000 companies are involved in a wide variety of industry scenarios, and the upstream and downstream ecological chains are complete, with huge resources for large-scale model training and testing.
The second is that DingTalk has obvious advantages in terms of user scale.
According to statistics, by the end of 2023, the number of DingTalk users has exceeded 700 million, the number of enterprise organizations has reached 25 million, and the number of software paying enterprises has reached 120,000.
Three years ago, the number of DingTalk users was about 400 million, why did it grow so fast?
Ye Jun said that many of DingTalk's new customers are upstream and downstream partners of existing customers, and the number of DingTalk users has achieved rapid growth in a short period of time in the form of customers leading customers. And this also shows that DingTalk has clear customer needs, compared with OpenAI's repeated "no-show" delay in launching the GTS store, the "privilege" in user needs is vividly reflected.
In addition, DingTalk has the "AI+" native reconstruction capability based on actual scenarios, and has iteratively upgraded the product many times in the past few months since it was connected to Tongyi Qianwen. At the same time, the general capabilities of AI that go deep into scenarios and real needs have been continuously strengthened and developed rapidly, which in turn has promoted the application of AI and formed a complete closed loop.
And this is in line with the characteristics that a super application should have as an entrance - a complete closed loop from technology to business.
In November last year, Alibaba announced the first batch of four strategic-level innovative businesses, and DingTalk was among them.
Wu Yongming, CEO of Alibaba Group, proposed that because of the advent of the AI era, DingTalk has gained unprecedented imagination. Every person and enterprise will have a personalized intelligent assistant, and DingTalk is expected to be the best AI intelligent assistant platform.
IDC's "**also**, AIGC will bring subversive changes to the form and format of application software. Minimalist interaction based on natural language will replace many traditional graphical interface interactions, forming a hybrid form of LUI+GUI. At the same time, the concept of "no app" will also reshape the next generation of applications, and various tools can be directly accessed and used through dialogue, so that more non-software professionals can also access powerful system services.
As a result, the super portal will become a typical front-end form for the new generation of application software.
Now, just two months after the strategy was announced, DingTalk has released 7Version 5 - AI Super Assistant, which concretizes ethereal imagination into clear operations.
Judging from the live demonstration and the audience's reaction, DingTalk's answer did not disappoint.
AI agent seems to be a complete attempt, and the concept of "no app" is beginning to take shape and land.
And along this path, the appearance of the new SaaS is becoming clearer and clearer.
Giant ERP will be broken up into more and more abundant small functions, and appear in the user's nearest position, "function to find people" instead of "people to find functions", fragmented characteristics more and more obvious;
The interaction is minimalist, and the human-computer collaboration is realized directly through the interaction mode of natural language such as dialogue, voice, and **, replacing the way people find the function menu.
At that time, based on the integrity of the interaction, product-to-business link and low barrier to entry, the "no SaaS" concept of software services will also be redefined.
When AI Agent becomes mainstream, anything is allowed to happen.