Smardaten's approach is to seamlessly integrate the robust data engineering capabilities of its proprietary "Data Pass" product, originally tailored for enterprise data services, with the comprehensive software engineering delivery process system of "Digital Pass," which caters to enterprise software services. Leveraging AI, they have successfully implemented these capabilities and systems within Smardaten 2.0, culminating in a sophisticated software development Agent product that facilitates end-to-end delivery for enterprises.

Author|Piye

Produced by|Industry Expert

In March of this year, Mu Hong was brimming with excitement.

Just days prior, DeepSeek emerged, instantly captivating the globe with its unique deep thinking model and open-source approach. People began to earnestly believe that "the era of AGI is truly upon us."

Long before this year, Mu Hong and the Smardaten he founded had been delving into the practical applications of AI. As a pioneering enterprise in China's low-code/no-code sector and a team deeply entrenched in big data for years, they were relentless in their quest to find an AI approach that could genuinely be applied in enterprise-level scenarios.

Their organizational structure mirrored this determination. Within Smardaten, the department responsible for future product pre-research and the experience department operate independently from the R&D department. Mu Hong's directive to these teams was to "lead the development of the next generation of products, not merely meet current customer needs."

DeepSeek has profoundly inspired Mu Hong and his team. "Previously, there was no method for deep thinking; most approaches relied on preset/building steps to generate results, which fell short of the intelligence definition," Mu Hong shared with us. "But DeepSeek's deep thinking model can overcome these challenges."

Driven by his vision, a series of innovative R&D initiatives have been vigorously pursued within Smardaten.

From an industry perspective, this move might not seem "swift." In recent years, amidst the AI model wave, numerous low-code/no-code vendors have launched their AI products, either focusing on the Agent level or integrating AI with low-code underlying engineering.

However, contrary to this enthusiasm, another fact looms large: for enterprises, truly usable and readily accessible AI products and components are scarce. Even fewer meet the criteria when adapted to the actual software development process.

"In fact, we developed the Agent platform early last year, but we withheld its release because we felt that technology wasn't the paramount factor; scenarios and applications, and a truly controllable end-to-end Agent product, were more crucial. The entire development process is precisely what we can control, and we believed we should first perfect the software development Agent," Mu Hong told Industry Expert.

This response aligns perfectly with Smardaten's latest move. Smardaten recently officially unveiled the Smardaten 2.0 version. In this upgraded version, a suite of AI Agent capabilities is showcased on the latest product interface. Based on the Smardaten 2.0 product, enterprises can not only swiftly generate software applications but also conveniently modify them through Agents post-generation.

"The enterprise-level implementation of AI is no mean feat; it's a systematic endeavor," Mu Hong emphasized.

As the industry delves deeper into the digital realm in 2025, and with high expectations for the value of large AI models, Mu Hong and Smardaten are delivering their answers.

I. Amidst the Industrial Evolution Tide:

Visible New Opportunities, Visible Digital "Old Bottlenecks"

Before dissecting Smardaten's response, a pertinent question arises: Where exactly has the wave of industrial digitization of Chinese enterprises reached? Or, in the past few years, amidst the AI model surge, what is the genuine pulse of digital intelligence?

According to a KPMG report, "Global AI Trust, Attitudes, and Applications Survey Report (2025)," over the past few years, more than 90% of leading enterprises in China have incorporated AI into their core strategies, and the usage rate of AI tools among small and medium-sized enterprises has surged by 31% year-on-year in the past two years.

Behind these figures lies a common aspiration of many large and even small and medium-sized enterprises - to elevate their digitalization level through AI. But can the AI available in the market truly deliver on this promise?

According to incomplete statistics, among all enterprises that have implemented AI, only 35% of AI projects can achieve large-scale application, with most enterprises still in the pilot phase. Software development is fraught with issues, such as insufficient data infrastructure and unclear links from demand to product design.

These are precisely the persistent problems that have plagued the software development customization process for many years. Whether it's demand research and scheme design in the initial stage, specific development and testing in the middle, or operation and maintenance in the latter stages, there are too many "leak points" for enterprises.

"Initially, we used Agents with the intention of leveraging AI to resolve these issues, and we also sought to adapt solutions from some low-code/no-code vendors, but the results were underwhelming," a digitalization leader from a central state-owned enterprise told Industry Expert.

This accurately reflects the current reality. Over the past few years, amidst the numerous AI service transformations, the low-code/no-code track has garnered significant attention. The rationale behind this is that many enterprises are striving to further ease the complexity of their software construction through the "low-code + AI" approach.

However, the truth is that, from the perspective of "low-code + AI" products available in the market, few enterprises have undergone such a substantial AI transformation - fully embedding AI into the development logic of low-code/no-code, which transcends being a mere question-and-answer assistant and necessitates true interactive operation Agents.

"Many low-code enterprises currently release products that primarily focus on the AI implementation stage, such as simple application frameworks and pages, without support for complex pages, complex charts, etc. This is because many platforms themselves are incapable of developing complex applications and do not support diverse customization functions," Mu Hong explained.

The challenge is not hard to foresee. To genuinely embed AI into the entire low-code development process, service providers must do more than just orchestrate and schedule inherent low-code component modules. They also need to engage in specific deep thinking tailored to different mature scenarios, such as business management and decision analysis, to accomplish complex tasks through the collaboration of multiple Agents.

Moreover, beyond these considerations, unlike the previous mode of independently invoking software and data systems, if service providers aim to achieve true AI software development delivery, a crucial link they must complete is to attain centralized management of "software + data," deeply integrating data models and software processes, and ultimately achieving unified invocation and implementation based on AI. Smardaten refers to this as "integration of data and application."

This is a sufficiently refined, industrialized, and long-term systematic project.

II. An AI Answer Sheet for "Software Engineering + Data Engineering"

The narrative returns to the beginning of this year. After swiftly aligning the team, Mu Hong and his colleagues began contemplating the key proposition he had earlier mentioned - "How to create an end-to-end Agent product?"

Based on this proposition, Mu Hong set more refined targets for the team. "Our initial objective was to 'increase application software delivery efficiency by five times.'"

The challenge lies in the fact that the Smardaten team not only needed to complete a series of Agent training and construction links at the technical level, such as data cleaning and annotation, knowledge base construction, and prompt engineering related to deep thinking, but also required the team to have a profoundly clear understanding of the various links of software delivery and their processes in diverse specialized scenarios.

However, this is precisely the core strength of Smardaten's years of expertise. Relying on the in-depth practice of the Smardaten platform, Smardaten has already disassembled and scenario-based precipitated the entire software delivery process - from demand research to development and deployment, and from general scenarios to industry-specific process rule sorting. These experiences have been transformed into a standardized process framework and scenario-based knowledge base embedded within the platform, providing underlying support for Agents' "deep understanding" of software delivery links.

Simultaneously, the two previously developed solutions, Digital Pass and Data Pass, form a robust complementarity: Digital Pass provides clear process navigation for Agents by constructing a standardized development link; while Data Pass's expertise in data governance and system construction lays a solid data foundation for technical links such as high-quality data annotation and scenario-based data modeling required for Agent training. This dual accumulation of "process experience + data capability" grants Smardaten a natural first-mover advantage when overcoming the obstacles in the implementation of Agent products.

With the synergy of these three products, according to public data, Smardaten now serves over 500 large enterprises and has amassed more than 1,500 components and 150,000 configuration items.

It can also be understood that these industrial know-hows, from atomized platform components to data system understanding and then to the full lifecycle of software development delivery processes, are precisely the confidence Mu Hong holds in building an end-to-end software delivery Agent product.

"We initially identified 104 main issues that hinder delivery efficiency and ultimately selected 10 of the most critical issues to focus on refining scenarios and utilize AI to address these challenges," Mu Hong stated.

The challenges also extend to the data front. Mu Hong informed us that to ensure the effectiveness of the end-to-end software delivery Agent product, the Smardaten team has thoroughly re-sorted through materials spanning the past decade. The delivery team, marketing department, and R&D department have converted these materials into Markdown text, which is easier for AI to comprehend, re-annotated them, and created a new knowledge base foundation.

These collective efforts have culminated in the now prominent Smardaten 2.0. In this upgraded version, AI occupies the most prominent position.

Specifically, the two focal points of Smardaten 2.0's AI capabilities are the intellectualization of software engineering and data engineering. Among them, in the software delivery link, enterprise developers can engage in software design for corresponding scenarios such as smart parks and industrial manufacturing through natural language interaction. From software data model design, architecture design to interface design, as well as adjustment and optimization of corresponding specific components, and subsequent visual analysis, data dashboard generation, etc., can all be managed through dialogue with intelligent generation assistants, significantly enhancing work efficiency.

On the other hand, the intellectualization of data engineering is oriented towards the internal data engineering system of the enterprise. That is, based on the relevant data agent capabilities of Smardaten 2.0, enterprises can intelligently optimize their data systems, from data cleaning to data annotation, data model construction to analysis report generation, helping enterprises cultivate a data environment conducive for AI to exert its maximum potential.

This aligns with another observation by Mu Hong and his team. That is, with the explosion of DeepSeek this year, although many enterprises, including central state-owned enterprises, have expressed clear intentions for AI implementation, there are still numerous bottlenecks in the execution process. The most prevalent issue is the imperfect data support system.

In Mu Hong's perspective, if Smardaten's prior role was more as a software development foundation, enabling enterprises to develop and customize personalized software based on its extensive low-code/no-code components on the platform, then in the current Smardaten 2.0, in addition to the inherent software-side AI upgrades, data or knowledge accumulation is even more pivotal for the future development of product competitiveness. Smardaten 2.0 transforms into the foundation for AI applications.

Smardaten's approach is to seamlessly integrate the robust data engineering capabilities of its proprietary "Data Pass" product, originally tailored for enterprise data services, with the comprehensive software engineering delivery process system of "Digital Pass," which caters to enterprise software services. Leveraging AI, they have successfully implemented these capabilities and systems within Smardaten 2.0, culminating in a sophisticated software development Agent product that facilitates end-to-end delivery for enterprises.

In essence, while these components are presented in a modular format on the product interface, they have also been fully embedded into the entire end-to-end Smardaten 2.0 software AI development process. Whether it's software development requirements, process adjustments, specific module (component) definitions, application frameworks, or backend dashboard generation, visualization, etc., enterprises can directly utilize them based on AI.

Mu Hong also shared that in addition to the end-to-end Agent capabilities of Smardaten 2.0, it also provides enterprises with an Agent building platform, enabling them to customize AI capabilities based on their unique needs and construct their exclusive Agent products utilizing different components on the platform, tailored to their enterprise scenarios and requirements.

"Our existing Sales Pass, R&D Pass, Service Pass, and over a dozen other internal applications are all powered by Smardaten 2.0. We maintain a seamless data link and scenario-specific data, enhancing these applications through Agents to bolster company efficiency and efficacy. Our initial in-house trial has equipped us to offer superior, more practical products to our clients," Mu Hong elaborated.

III. Navigating the Depths of Industrial Digitization

What should an enterprise's AI infrastructure resemble?

If there is a dominant theme for 2025, AI undeniably tops the list; dissecting this further, the implementation of AI stands as the central focus.

Behind this theme lies not just a refined emphasis on expressing AI technology's value but, crucially, in the current landscape where enterprise digitization has plunged into deeper waters, the AI capabilities and foundations enterprises must rely on to transcend the constraints and challenges of digital transformation.

Smardaten 2.0 has been fully integrated within Smardaten, yielding exceptional results across various software development stages. For instance, during the demand research phase, leveraging intelligent assistants boosts software prototype drawing efficiency by 80%, significantly reducing the demand confirmation cycle for a moderately complex functional module from an average of 3 days to merely 4 hours.

Another example is the scheme design phase, where system design engineers utilize AI to retrieve knowledge bases, securing mature business design schemes that can be delivered after minor adjustments tailored to project business scenarios, decreasing the overall design workload by over 60%.

"We prioritize using the product internally, crafting solutions for each link with a clear data loop using Smardaten 2.0. Only after rigorous internal testing can we confidently offer it to enterprise clients," Mu Hong stressed. "Subsequently, we will gradually open it up for select clients to trial, providing an official version to enterprise clients only after ensuring outstanding results."

From the lens of AI implementation and digital transformation, Smardaten 2.0 equips enterprises with a foundational AI capability. Built on this foundation, enterprises can advance the software development process more swiftly, efficiently, and intelligently, adapting specialized AI digital transformations tailored to different scenarios, integrated with their accumulated, specific knowledge bases.

Mu Hong further emphasized that beyond AI product capabilities, enterprises must also undergo organizational transformations. "For example, software development based on Smardaten 2.0 differs significantly from previous processes, altering interdepartmental relationships and collaboration methods." These changes are gradually being implemented within Dataway Data.

There is a consensus that as enterprises delve deeper into digital transformation, they require a new foundation for the AI era. This foundation must encompass inherent low-threshold development capabilities like no-code/low-code, along with a suite of AI capabilities such as data intelligence and delivery intelligence. This enables enterprises to construct AI applications across various links with minimal thresholds and maximum cost-effectiveness, further addressing inherent digital bottlenecks and challenges through AI, thereby enhancing business competitiveness.

This is precisely Dataway Data's mission. It is understood that an internal roadmap envisages Smardaten 2.0 becoming Dataway Data's primary R&D version over the next two years. Concurrently, adhering to the principle of "delivering one generation, researching and developing one generation, and pre-researching one generation," the 3.0 version, focusing on Agentic Agent capabilities, is also under development.

Starting from practical challenges, dismantling AI models through a "technology + industry" perspective, progressing from self-validation to industrial empowerment, this is the genuine AI productivity path explored by Mu Hong and Dataway Data over the past few years.

"Some view Agents as a new enterprise service paradigm. I believe that for Agents to truly take root in the enterprise service domain on a large scale, they must be code-free. Our goal remains unchanged: to empower everyone to develop Agents," Mu Hong concluded.