Source: AI Whale Selection Society

Image Source: Generated by Wujie AI

"The number of DingTalk app users has exceeded 10 million, but by the end of 2020, it was still less than 1 million," exclaimed DingTalk President Ye Jun (nickname Buqiong).

So many software applications operate on enterprise office platforms, carrying multiple contents such as enterprise information exchange, operational management, and business operations, making it difficult for the platform to "slim down." Therefore, enterprise office software is all moving towards the direction of becoming "all-encompassing."

For example, in order to ensure that information reaches its destination, DingTalk developed "Ding a Message"; in order to allow documents to be used by more people simultaneously, Feishu opened up document collaboration to 5,000 people; in order to increase revenue, WeChat Work started charging for contact lists exceeding 2,000 people. "All-encompassing functionality, all-encompassing capabilities, all-encompassing commercialization" can be said to be the three "all-encompassing" aspects of enterprise office products.

However, as development progresses, people are gradually realizing that both internal and external "all-encompassing" approaches make it difficult to achieve a win-win situation. For example, DingTalk's live streaming capability experienced significant development during the pandemic, but it also led to jokes about elementary school students giving it a one-star rating; Feishu's pixel-level feature development led to a team size of up to 6,000 people; the commercialization of WeChat Work also made many small and medium-sized enterprises feel overwhelmed.

In order to address these unavoidable development issues, DingTalk proposed the low-code (aPaaS) revolution in 2021, and from then on, the gears of destiny began to turn. Successively developing foundational capabilities such as bPaaS, iPaaS, and dPaaS, DingTalk built a foundation and framework through PaaS, creating various functional services by leveraging the modular building blocks of ecological enterprises.

DingTalk is no longer engaging in "all-encompassing" industry practices, but is raising the PaaS platform and open strategy to build an ecosystem. This means that DingTalk only provides basic products such as documents, audio and video, projects, and meetings, while leaving everything else to the ecosystem, including industry applications, professional applications for personnel, finance, materials, production, supply, and sales, and research.

Even greater change came in 2023, when DingTalk announced the integration of the Tongyi large model, giving DingTalk the ability to output AI. From the slash function in June to the release of AI PaaS capabilities on August 22, and now the emergence of concrete AI applications and functions such as digital employees and AI assistants.

"DingTalk has only a little over 1,500 employees, but there are 150,000 service providers, each city has independent service providers. Together, we serve 23 million enterprise organizations and 600 million users," Buqiong said in a group interview with AI Whale Selection Society and other media.

DingTalk's development path is gradually becoming clear. In the wave of the AI era, can DingTalk become the biggest variable?

DingTalk Becomes Cool, Large Models Bring AI Wind

"DingTalk is becoming cooler, especially after the second wave of AI office," this is the evaluation of many live broadcast bullet comments.

At the Spring Ding Summit on April 18, DingTalk released a "slash", which is the first wave of AI for DingTalk and even the entire collaborative office. It is understood that the slash function was already launched in 2021, but due to its simplicity, few people used it. With the AI wind brought by large models, the slash's capabilities have become increasingly powerful and are now being liked by users.

Today's slash can already serve as a full-time assistant for enterprise employees. From the on-site demonstration, DingTalk's slash can create a clear mind map with just one prompt sentence, "Please generate a sales plan for smartwatches during the Double 11 period." It covers product introduction, target market, competitor analysis, sales plan, and profit expectations. Managers can use this as a basis to create more comprehensive execution plans based on their company's specific situation. Subsequent product e-commerce promotional posters, commercial analysis reports, and after-sales robots can also be quickly generated with AI involvement.

These capabilities go beyond GPT-like products and are more adept at writing promotional copy, gradually landing in real business scenarios for enterprises. Many functional components have been directly turned into services through AI, improving the efficiency of enterprise office work and reducing the level of internal competition among employees in the workplace.

DingTalk's personal version, which started internal testing in August, is responsible for Notion+ChatGPT+Midjourney+Zoom, according to DingTalk's personal version leader Ma Ruila in a post on Jike. Through AI, personal work efficiency is improved, and DingTalk's progress even rivals Microsoft's Copilot.

At the same time, DingTalk's AI PaaS allows the industry ecosystem to no longer be internally competitive at the foundational level.

Initially, DingTalk greatly reduced the development difficulty of many applications with low-code, and currently, the number of low-code applications in DingTalk has exceeded 8 million. Now, AI further reduces the development threshold, transitioning application development from dragging and dropping functional modules to directly generating applications using natural language.

The digital employee "Recruitment Specialist (1haozhaizhai)" co-created by DingTalk and 1haozhipin is an example of this. This digital employee can help HR departments write job postings, post them on relevant job sites, collect resumes, screen resumes, and schedule interviews.

The magical journey begins with a prompt sentence, "Help me write a job description for a senior product manager," and 1haozhaizhai will immediately generate it within one minute, saving time for HR to write job postings. For HR, the most time-consuming task is resume screening, and 1haozhaizhai can use AI to compare resumes and extract the strengths and weaknesses of two candidates directly. If HR values "5 years of relevant industry experience" and "good education," suitable candidates can be quickly screened and invited for interviews.

According to DingTalk CPO Qi Junsheng, DingTalk's digital employees can enter the enterprise directory and obtain corresponding permissions based on different roles, greatly improving the applicability of digital employees.

Previously, developers had to create various capabilities, but with the integration of large models in DingTalk, they only need to specialize in tuning AI applications and focus on creating popular applications.

For the platform, DingTalk has been pushing for the infrastructure construction of the open platform, and now it is gradually showing its accumulated advantages.

Previously, DingTalk opened a large number of API interfaces, plug-in systems, and RPA systems. The release of bPaaS and "Cool Apps" in 2022 integrated business applications into high-frequency scenarios such as group chats, laying a certain foundation for the entry of AI capabilities into high-frequency scenarios.

Now, more fun AI products are being co-created by the ecosystem.

In DingTalk's audio and video conferences, the "digital avatar" can attend the meeting on behalf of the user, help the user understand the important information of the meeting, and push the meeting summary afterwards;

DingTalk, in collaboration with Ele.me, created "Happy Group Purchase," as demonstrated by Buqiong. Users can place orders and group purchases in DingTalk groups using natural language and make a single payment. 30 minutes later, 20 cups of coffee will be delivered to the venue by a delivery rider in a blue uniform.

"AI Assistant" can help teachers intelligently grade assignments, analyze student learning data, generate teaching suggestions, and create PPTs for explanations.

Just as two years ago, DingTalk did not anticipate that low-code would rapidly expand Ding applications to tens of millions, now that AI capabilities have been opened, what fun products will Ding applications give birth to, and how quickly they will expand, are all unknowns.

Open to the End, A Fresh Move Breaks Through the Sky

DingTalk's initial dream was to reach 1 billion users. However, for DingTalk, focusing only on information exchange has limited value within enterprises. To make enterprises enjoy using it and to increase the value of To B, DingTalk has decided to go further with the ecosystem.

Therefore, on March 22, 2022, at the 2022 Ecological Conference, DingTalk announced the opening of the ecosystem and proposed the "PaaS First Partner First" strategy, clearly stating that "DingTalk only does one thing, which is PaaS-ification." It is this move that has made DingTalk's future strategies different.

Buqiong explained the view of the ecosystem in this way. "A cake can be cut from top to bottom, or it can be cut horizontally." Many platforms compete with SaaS companies, "doing everything from start to finish," which is a vertical cut. The PaaS approach turns DingTalk into a "horizontal cut".

"Alibaba Cloud made a horizontal cut, and we also made a horizontal cut, leaving other tasks such as spreading cream, decorating, and adding fruit to the companies in the ecosystem," explained Buqiong. Whether ecosystem companies are willing to develop on DingTalk's enterprise depends on DingTalk's platform being able to output stronger capabilities, and AI PaaS is DingTalk's strongest output capability.

DingTalk's AI PaaS, from the underlying model training platform and plug-in platform to the middle model scheduling platform, and then to AI applications such as digital employees, has built a relatively complete output framework, making it relatively easy for developers to create various applications.

It is reported that since the invitation test was launched in June, more than 100,000 customers and ecosystem partners have enabled "slash."

On August 22, at the 2023 DingTalk Ecological Conference, DingTalk President Ye Jun announced the latest progress of DingTalk's comprehensive intelligence: 17 product lines and 55 scenarios have been fully integrated with large models, completing intelligent reconstruction; DingTalk has also opened up intelligent base AI PaaS to ecosystem partners and customers, not only having Alibaba's Tongyi Qianwen, but also being able to accept other large models such as Wenzin and Xinghuo.

With more and more AI applications such as 1haozhaizhai, DingTalk has launched an AI intelligent application market, and the first batch of 11 intelligent SaaS applications have been launched. Currently, low-code applications account for 80% of Ding applications. When AI intelligent applications account for this proportion, intelligent applications will emerge, solving the biggest challenge facing Chinese enterprises and SaaS.

This challenge is the contradiction between customizing customer demands and standardizing product scale for profitability. "B enterprises will definitely not use A enterprise's solution, and proposing this demand, changing this feature, will continue to increase costs," DingTalk makes it easier for enterprises to deliver through the capabilities of the open platform.

Over the past year, DingTalk has gradually explored a sustainable commercialization path - PLG (Product-driven and PaaS-driven base) + SLG (Sales-driven and Service-driven). Simply put, in the past, the digital transformation of large enterprises required the deployment of hundreds of people for about a year to deliver. When DingTalk has the capabilities of PLG and SLG, it can first output to service providers, and the service providers can then help enterprises with digital transformation using DingTalk's capabilities.

For example, the large customer serviced by DingTalk, FAW, directly cooperates with a technology company "FAW Qiming" under FAW. This software service company has more manpower than DingTalk and has a better understanding of FAW. DingTalk gives the capabilities to them instead of directly serving FAW employees. This has improved efficiency, and the ecosystem service providers have also made money.

Openness is not just a strategy, it is more of a capability. "If we and the service providers don't serve a company well, that company will not cooperate with us for at least 3-5 years," emphasized Buqiong. To B services are not a one-time deal, and DingTalk has experimented for a long time to verify that this model can meet user needs.

Stopping the bleeding, AI opens a new door

SaaS companies, including DingTalk, have not been profitable for a long time, which is a common problem in the Chinese SaaS industry.

The reasons are multifaceted, one of which is high costs, not only in customized delivery, leading to SaaS losing marginal effects. At the same time, server and bandwidth costs are also frighteningly high.

It is understood that after the pandemic, as DingTalk entered normal usage, server and bandwidth costs have dropped to one-fourth of the peak of the pandemic. From Alibaba's latest financial report, it can be seen that in the second quarter of this year, Alibaba Cloud's adjusted EBITA was 387 million yuan, a year-on-year increase of 106%, mainly due to the normal usage of DingTalk compared to the same period last year, leading to a reduction in hosting and bandwidth costs.

Another reason for not being profitable is the difficulty of commercializing SaaS. The "2021 China SaaS Survey Report" shows that the paid rate for office software in China was less than 15% in 2021, while in the European and American markets, the paid rate for office software has exceeded 70%.

The reason for the huge gap is that the current acceptance of SaaS by Chinese users is shallow, and DingTalk, relying on the accumulation of a large number of small and medium-sized enterprise customers for free, has low order amounts and weak payment awareness. In September of last year, Ye Jun mentioned, "Now companies basically use DingTalk for free, and the number of truly paying customers is probably less than 1%."

Previously, DingTalk's intelligent hardware revenue once accounted for over 80%, but now it has dropped to below 10%. "Commercialization is not achieved overnight, and we have tried many paths," Buqiong said. Before DingTalk took the open path, it also experimented with many commercial projects.

But now, DingTalk proudly announced commercialization data. As of the end of March 2023, the number of paid enterprise users of DingTalk software reached 100,000, with small and micro-enterprises accounting for 58%, medium-sized enterprises accounting for 30%, and large enterprises accounting for 12%; DingTalk's paid DAU exceeded 23 million.

Buqiong revealed that DingTalk's ARR (Annual Recurring Revenue) far exceeds the $100 million standard of a unicorn company, and the GAAP annual subscription revenue has also significantly exceeded this figure. ARR reflects the signing status of customers for the year and is a basic indicator of the operating performance of SaaS software companies; GAAP revenue reflects the actual performance for the year, more truthfully reflecting the actual operating situation.

It is understood that the ARR index announced by DingTalk only includes software subscription revenue, excluding project-based, hardware, API, and other revenue, so it has a good value. This also means that the current situation of unprofitable Chinese SaaS is being changed by DingTalk.

DingTalk is also trying to open up more commercial capabilities. The latest news is that DingTalk has provided the industry's first commercialization solution for large model landing application scenarios: on top of the annual fee of 9,800 yuan for the DingTalk professional version, an additional 10,000 yuan can be added to get 200,000 calls to the large model; on top of the exclusive DingTalk annual fee, an additional 20,000 yuan can be added to get 450,000 calls to the large model. This means that the average cost of a single call is less than 5 cents.

However, Buqiong also emphasized that they are currently in the stage of growing the ecosystem and are not in a hurry to make money. For example, Chongqing Dianming Technology, founded by Chen Yaojian, as a service provider for DingTalk, achieved revenue of over ten million yuan in the first year of entrepreneurship. "Sometimes they serve a customer with a dedicated version, and they earn over 100,000 yuan, while we earn 9,800 yuan, and there is also their commission in it."

According to the data released, as of March 31 this year, DingTalk has over 5,000 ecosystem partners, including independent software developers (ISVs), consulting ecosystems, sales and delivery service providers, and hardware ecosystem manufacturers. The number of partners with annual revenue exceeding ten million yuan reached 25, and for every yuan of revenue DingTalk earns, it brings nine yuan to its ecosystem partners.

By allowing ecosystem partners to make money, DingTalk focuses on platform capability output and ultimately, with the landing of AI products, DingTalk will not become the next powerful single-domain Salesforce, nor will it become a market for selling third-party applications, but will become an AI open platform.

DingTalk insists on the path of growing the ecosystem while continuing to maintain the attributes of a technology platform, which also maintains its imagination for the future when facing the capital market after its recent independence from Alibaba Cloud.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。