What did those companies that were the first to implement AI do right?

Source: Geek Park

Author: Su Zihua

In the past year, the term AI has been almost everywhere in the business world.

Some companies have allocated hundreds of thousands to millions in AI funding budgets at the beginning of the year; some executives are busy holding AI strategy meetings; and others have formed special AI teams...

From last year's hesitation and wait-and-see attitude to this year's proactive layout, Shen Tao, Vice President of Strategy at Fanruan Software, stated: "Last year, it might take 3 months to knock on the customer's door, but this year, after the Spring Festival, customers took the initiative to reach out. This is a tremendous change."

Behind this, the B-end AI implementation has迎来了真正的历史性机遇.

But in the end, we often hear feedback like: "The technology is there, but why isn't it being used well?" "The actual effects are not being realized in the industry." - Many AI projects have not really been implemented.

Investment is real money, and anxiety is also very real.

The contradiction lies in the disconnection between technology and application scenarios. Many business managers have reported that AI products perform excellently in demonstration environments, but frequently "crash" in real business situations. This contrast between the "demo myth" and the "implementation dilemma" exposes the limitations of companies going it alone—either they lack strong foundational model support, or they find it difficult to transform general technology into industry-specific solutions.

So, what did those companies that have truly implemented AI and achieved commercialization do right? After chatting with leading players in the industry such as Tuya Smart, Fanruan, Lanling, and Autonavi, we found that the key to success is to carve out a path together with cloud platforms— a new route from technology to scenarios and then to commercialization.

Their "AI landing" results indicate that the industrial implementation of large models requires teams that focus on vertical scenarios to collaborate with cloud platforms to build a co-creation ecosystem for AI products, allowing technology to truly integrate into enterprise processes and products, rather than just achieving isolated breakthroughs.

01 Co-building AI has expanded the business boundaries of enterprises.

A technology shifts from hype to value, the key is "who can use it."

In the past year, companies that have truly implemented AI applications share a commonality: they do not fight alone, but rather "co-build" with cloud platforms. Everyone realizes that in the rapidly changing AI industry environment, collaboration is the most efficient survival strategy.

In the past, cloud vendors provided model APIs that enterprises could integrate; however, the logic has changed now. For example, in the AI ecosystem co-built by Alibaba Cloud and industry partners, Alibaba Cloud actively participates in the product co-creation process: from defining scenarios, packaging components, integrating data, to supporting the commercial pathway. The role of cloud vendors is evolving from infrastructure providers to value co-creation partners.

This co-creation is not just "you use my model", but rather "we define the product together". Ke Dumin, Vice President of Tuya Smart Technology, stated that in creating the "Tuya IoT Platform Alibaba Cloud Version", "we co-created this product with the Alibaba Cloud marketplace, visiting customers together, understanding their needs, and defining the product together."

"Tuya IoT Platform Alibaba Cloud Edition" can help industrial clients' devices go to the cloud and implement AI capabilities. Tuya Smart Technology Vice President Ke Dumin revealed that they initially approached it with a trial attitude but unexpectedly gained numerous commercial clients.

Therefore, the essence of co-creation is to define the incremental market together, making cross-border innovation possible. The effect of one plus one being greater than two becomes evident at this time, where Tuya Smart has expanded its business from focusing on smart space scenarios to multiple new fields such as agriculture, retail, and manufacturing, successfully implementing the world's top-ranked livestock intelligent management project in Singapore; meanwhile, Alibaba Cloud, which provides AI technology and cloud services, has also expanded into new markets.

Ke Doumin stated to Geek Park: "With the arrival of AI, many industries are worth redoing. Industries like emotional companion toys and consumer-grade headphones had little connection with IoT in the past; however, now, large models need the collaborative support of IoT technology to transition from the digital world to the physical world." He further pointed out that the emergence of large models not only opens up new growth opportunities for these industries but also further strengthens Tuya Smart's existing business advantages.

As a company that started with smart home solutions and gradually expanded from indoor to outdoor AIoT platforms, Tuya Smart is leveraging large model technology to drive each IoT product to integrate AI functions and attributes, matching corresponding application scenarios—from single device smart upgrades to "spatial intelligence." Ke Dumin mentioned that the AI-driven "home brain" will be able to more effectively enhance user experience and the level of scene intelligence.

Similarly, after Fanruan launched the Tongyi Qianwen plugin on its Jiandaoyun platform, they did not do any complicated packaging and found that customers began to use it automatically. Shen Tao admitted: "We didn't specially design any scenarios; we just launched the plugin, and as a result, customers started using it themselves."

It can be seen that tools with low barriers to entry and high adaptability can best stimulate users' real needs. In the daily business handled by Jiandaoyun, AI plugins have played a key role in scenarios such as contract review, resume screening, and customer follow-up analysis. Customers no longer need contract reviewers with a monthly salary of five to six thousand, nor do they need to manually sift through customer records to extract demands—AI can automatically identify key information such as signing intentions and price fluctuations.

In the case of large enterprises, the power and effects of co-construction are even more significant. Blue凌, which specializes in serving central state-owned enterprises and large corporations, has upgraded their "Blue Doctor" from being an intelligent Q&A product within the company to an "AI Middle Platform" through large models and toolchains.

Built on the framework of "General Knowledge Q&A + Exclusive Small Model + Intelligent Agent", the new "Dr. Blue" not only provides intelligent Q&A but also enables cross-system search, experience extraction, and the AI transformation of applications such as document completion and processes.

After landing on the platform, Blue凌's first new energy customer, Silis, achieved the "three ones": finding work knowledge in one minute, initially solving problems in one day, and accumulating project experience in one month.

The exponential improvement in efficiency is the most direct contribution of AI to enterprises.

The achievements of the collaboration between Lanling and the cloud platform indicate that AI capabilities must be transformed into usable products for customers, and both the platform and industry know-how are indispensable. "Alibaba Cloud has the technology and customer resources, but many concrete scenarios require our implementation," said Xia Jinghua, director of the Lanling Research Institute. "We need to work together on this."

A more typical example is the MCP service of the Gaode Open Platform. By overlaying the semantic understanding of Tongyi Qianwen with its own mapping capabilities, developers can generate a complete cycling route with just one natural language sentence, and automatically generate map code.

This "model + MCP + toolchain" approach has greatly expanded Gaode's business boundaries and created new business opportunities for developers. A relevant person in charge at Gaode stated to Geek Park: "The introduction of large models can better help our services upgrade from a single map to an all-scenario travel solution. We hope to reach more customers through the ecosystem."

Through the numerous cases above, we can see that the boundaries of enterprises are being redefined. They will not only be determined by labels such as industry and scale but also by "what problems can be solved." In the process of co-building AI, industry partners are able to break through their own limitations and enter fields that were previously difficult to reach.

For cloud platforms, the process of co-building the AI ecosystem also promotes their transformation from "selling capabilities" to "ecosystem organizers". It can be said that the breadth of the platform's technology and the depth of the industry partners' scenarios form the golden combination for the implementation of AI.

02 AI Commercialization: Entering the Ecosystem Competition Stage

Two years ago, when large models were just emerging, companies were still competing on parameters and fighting their own battles. As we enter 2025, the industry is increasingly focused on the real issue of "how AI can be monetized."

In the past, the frequently mentioned term was "model performance"; now, what appears more often are "scenario-based Agent", "deliverable solutions", and "channel monetization".

The cases of Fanruan, Lanling, Tuyuan, and Gaode indicate that in the co-construction of the "AI ecosystem" with partners such as cloud platforms, what is built is not only the technology stack and product capabilities, but also the commercial channels. The core value of the ecosystem lies in bridging the "last mile" from technology to business.

For example, Lanling leverages Alibaba Cloud's customer resources and market subsidies to acquire new customers and expand overseas; Gaode Open Platform will soon launch the Gaode MCP Server on the Alibaba Cloud marketplace, directly connecting to the developer ecosystem; Fanruan has revealed that they are trying to co-create the Agent solution with Alibaba Cloud, listing it on the Alibaba Cloud marketplace and converting platform traffic into commercial results.

As leading companies accelerate monetization through ecosystems, industry analysts predict that by 2030, 50% of enterprise AI models will be privatized domain models, while this proportion was only 5% in 2024. This means that the future implementation of AI will rely more on the close collaboration between "general large models + industry small models + scenario-based tools."

These business moves reflect a change and trend: the implementation of AI is a systematic project, and platforms need to provide end-to-end support. Enterprises' expectations for cloud platforms are no longer solely focused on model performance; they are beginning to hope that platforms can provide product delivery capabilities, market reach capabilities, and even joint operation capabilities.

As the saying goes, technology determines the lower limit, while the prosperity of the ecosystem will determine the upper limit. In April of this year, Alibaba Cloud's "Blossom Plan" is precisely a footnote to this transformation.

According to the official definition, the "Blooming Plan" aims to make strides in six key areas over the next three years: infrastructure, models, data, tools, applications, and delivery, with the goal of serving one million customers and achieving a hundred billion in business together with partners.

The cases of Fanruan, Gaode, Tuya Smart, and Lanling, which we mentioned earlier as having made good progress in the implementation of AI, are precisely the co-building partners of the "Flourishing Plan."

From an external perspective, the "Blossom Plan" reveals a subtle transformation in the role of Alibaba Cloud. It can be likened to building a shopping mall; in the past, it was only responsible for constructing the building and providing electricity; now, it must attract various merchants, assist restaurants in designing menus, help clothing stores set up display racks, and even coordinate supply between merchants.

The value of the "Flourishing Plan" lies in the current anticipation across various industries for the implementation of AI applications. It has fostered an ecological system that enables lower friction in collaboration and higher density in innovation. Reducing the costs of ecological collaboration and enhancing innovation efficiency will become the core competitiveness of the platform.

In this ecosystem built by Alibaba Cloud and its partners:

Openness is the cornerstone of ecological prosperity. The cloud platform provides a truly open ecosystem through open models, data, toolchains, cloud markets, etc.

Ecological partners transform industry know-how into replicable product solutions;

Market channels and business mechanisms support the business closed-loop transformation "from proposal to signing."

The ultimate goal is for participants to collaboratively promote the "demo showcase" into a "real application."

Alibaba Cloud's initiatives in the product ecosystem dimension also provide us with an insight: whether now or in the future, the winners of the AI era will be those who find the right partners, identify the right scenarios, and turn technology into usable products. Ultimately, by 2025, the competition in AI development will not just be about "whose technology is more dazzling," but rather "whose ecosystem can deliver."

Perhaps this is also an extension of Alibaba's philosophy of "making it easy for everyone to do business" in the AI era - "making it easy for everyone to do AI business."