Lingyang released AgentOne. Being practical is more important than being "fast".
Halfway through the first year of the Agent era, how is the most "challenging" enterprise-level Agent faring?
An obvious trend is that not many fancy new concepts have emerged in this field. When it comes to the question of whether to implement Agents immediately, many enterprises are still hesitating and observing. After all, this is a group that won't pay for "showy technology."
However, at the same time, a crucial turning point is approaching.
Qianxun Spatial Intelligence, a spatio-temporal intelligent technology company that launched its first digital employee, officially issued the first AI "employee ID." Since collaborating with Lingyang to develop an intelligent customer service Agent in March this year, Qianxun Spatial Intelligence has been conducting long-term evaluations of the Agent.
The KPIs for the successful "full-time conversion" of the intelligent customer service include an 80% accuracy rate in answering questions and a 50% completeness rate in responses. That is to say, now half of the customer inquiries can be resolved independently without transferring to human customer service.
The high learning ability, the efficiency of the customer service playing multiple roles, and the "release of human efficiency" that encourages employees to actively promote AI iteration are all tangible benefits that Qianxun Spatial Intelligence has experienced during the implementation of the Agent.
In 2025, Qianxun Spatial Intelligence plans to continue collaborating with Lingyang and aims to incubate more than eight AI employees. This also shows that the current digital employees are not only usable but also becoming increasingly useful.
Although large models have strong capabilities, it is difficult to directly address the specific needs of enterprises. This deadlock is being broken as enterprise-level Agents are applied in more scenarios. The digital visions that were previously difficult to implement and slow to show results now have the realistic possibility of large-scale implementation for the first time.
"In the Agent era, you'll find that when the computing power cost of large models reaches a certain critical point, a leverage effect will occur, making enterprises willing to invest. But before that, judgment and perseverance are needed."
Lin Yongqin, the vice president of Lingyang, an intelligent business service company under Alibaba Cloud Intelligence Group, told 36Kr that this enterprise service company under Alibaba is ready to continue doing the "most difficult" things.
01. Why is it challenging to develop enterprise-level Agents?
What are the difficulties in developing enterprise-level Agents?
On the surface, it is more difficult than expected to encourage enterprises to shift from "passively accepting" to "actively embracing" Agents.
"Whether it was SaaS in the past or Agents today, the principles that enterprises follow when considering adoption and payment are actually the same. ToB Agents need to be deeply integrated into the decision-making and operational processes of enterprises. One shouldn't start blind trials just because the technical threshold is low."
Lin Yongqin, Vice President of Lingyang, Alibaba Cloud Intelligence Group
Lin Yongqin said that enterprises are more practical and calm than C-end users when it comes to new things. When new technologies emerge, enterprises are most concerned about three aspects: value, cost, and security.
The blurred boundaries in these three aspects have also led to many concerns among enterprises regarding the implementation of Agent applications.
Enterprises embrace Agents not to follow the technological trend. Their core requirement is to achieve stable and quantifiable efficiency improvements. Therefore, in Lingyang's view, enterprise-level Agents need to compete not only in technology but also in the selection and in-depth understanding of scenarios.
Since July this year, within more than two months, Lingyang has successively released three batches of enterprise-level Agents, covering high-demand and high-frequency scenarios such as customer service, data analysis, and marketing, and has also delivered tangible results.
At this year's Yunqi Conference, Lingyang also launched AgentOne, an enterprise-level AI intelligent agent service platform, upgrading the capabilities and system of Agents to the platform level.
As a digital and intelligent enterprise service provider under the Alibaba ecosystem, Lingyang's Agent products were not launched very early. Compared with general Agents that have been in the market for half a year, Lingyang's strategy of seizing the opportunity later comes from its determination to thoroughly understand and deeply explore scenarios.
At the end of 2023, the Lingyang team began to explore whether the capabilities of Agents based on foundational models could produce results in specific fields such as customer service and marketing. Through continuous trials and feedback, they started to build the platform.
It wasn't until the second half of this year that Agents gradually entered the mainstream view and had some valuable scenarios suitable for implementation. As customers' demand for Agents began to surge, Lingyang verified the implementation effect of its products through co-creation with individual customers. Therefore, it chose to officially launch its Agent products to seek more value verification.
"In the ToB field, what matters most is not who enters the market early or late, but the depth of product implementation. So when everyone was rushing for time in the early stage, we were still determining the most suitable time window for expansion."
Lin Yongqin believes that the reason why Lingyang can launch and iterate its Agent products at a steady pace is that in the early stage when the concept was just emerging and the value perception was still unclear, Lingyang, with its years of scenario insight and keen sense, chose to stand with enterprise customers. For example, it started by standardizing and refining private domain data, avoiding the security risks of enterprise Agent implementation.
If ToC Agents are like icing on the cake for content and office efficiency, ToB Agents are more like timely help. Letting AI optimize the inefficient links in enterprise operations is like "treating a disease." Only by removing the root cause can the enterprise gain more stable strength through different means.
These are also the deep-seated reasons why enterprise-level Agents have become the "toughest nuts to crack":
First, compared with C-end applications, enterprise scenarios have long business chains, fragmented scenarios, and too many non-standardized scenarios. A single Agent cannot solve all problems. For example, a complete customer service process may involve multiple breakpoints such as consultation, ordering, after-sales service, and logistics. If the Agent cannot be integrated with the existing business system to form a closed loop, its value cannot be significantly improved.
Second, the efficiency of data assets is low, and the cost of data structuring is high. Data is the fuel for Agents, but the data within enterprises is often scattered, isolated, and uneven. To make good use of Agents, enterprises must first invest a large amount of cost in data governance and integration, which is precisely the weakness of many enterprises.
Third, security and stability are the bottom lines for enterprises to accept the implementation of new technologies. In the past two years of AI deployment, many enterprises have found that although the demonstration effects of many Agent products are very impressive, they may encounter problems such as response delays, judgment errors, and system crashes in actual business scenarios. An unreliable Agent not only fails to improve efficiency but also brings business risks.
For these reasons, Lingyang has built AgentOne as the first stop for enterprises to implement AI applications. It provides a full-link development workspace for Agents, supporting the construction, evaluation, analysis, optimization, and deployment of Agents. By connecting with the existing enterprise systems, it can shorten the implementation cycle.
02. What does a "down-to-earth" Agent look like?
Why can't many Agent manufacturers solve the same problems well? What has Lingyang done right?
The competitive advantage of this company lies in moving beyond the technology-driven stage of "looking for nails with a hammer." By combining its industry know-how, the unique Alibaba ecosystem, and practical productization capabilities, it has formed a comprehensive solution that is difficult to replicate quickly.
In the view of Peng Xinyu, the CEO of Lingyang, Alibaba Cloud Intelligence Group, there is a golden formula for building enterprise-level Agents: large model × high-quality data × strong scenarios.
Peng Xinyu, CEO of Lingyang, Alibaba Cloud Intelligence Group
Therefore, those who can "read data well" have already made good use of half of AI. In terms of the team, Lingyang originated from Alibaba's data middle platform and grew up in the practice of enterprise data intelligence. It has a deep technical foundation for enterprise data.
As Peng Xinyu mentioned at the 2025 Yunqi Conference: "Enterprises having data doesn't mean they have high-quality data. Enterprise data is like building blocks. Without a blueprint, even with many blocks, you can't build any building. High-quality data represents structured and logical data."
The intelligent foundation must be rooted in high-quality data. So many enterprises have invested a lot of cost and effort in data governance to use AI. Lingyang, which has a strong foundation in data, is currently one of the few suppliers that can help enterprises overcome this challenge.
"It can be observed that the so-called 'dead data' in enterprises and some data previously considered to have low ROI have all received more value improvement, and can even be quantified and monetized."
Lin Yongqin said that previously, enterprises operating on multiple platforms only analyzed the raw data in pictures through a single platform. But after the introduction of Agents, these enterprises hope to have a unified processing format so that AI can better understand the data language.
In other words, a practical Agent can help enterprises pay more attention to the value of data.
In the data analysis scenario, Lingyang has launched a series of intelligent agents such as the Question Data Agent and the Interpretation Agent, which are currently being applied to hundreds of enterprise customers in different fields such as Muyuan Foodstuff. This is where the value potential of enterprise-level Agents lies.
Compared with many manufacturers that have transformed from software and productivity tools to Agent development, Lingyang is also regarded by many customers as a "down-to-earth" AI application provider: it doesn't overstate its capabilities or create unnecessary anxiety but focuses on solving the universal needs of "having data to rely on and having results to prove" in real business scenarios.
Secondly, Lingyang's in-depth understanding of scenarios provides specific "springboards" for data intelligence. This is what Peng Xinyu calls the "strong scenarios" required by enterprise-level Agents.
"Places with high labor density and high data density are where humans and AI can coexist and progress in the future." Peng Xinyu said that the reason for abstracting enterprise-level Agents into three key elements comes from Lingyang's extraction of common values from the complex business processes of many enterprise customers.
Take Fosun Tourism Group, a customer of Lingyang, as an example. The pain point this tourism and culture group faces is that its service chain is extremely long, customer needs are highly personalized, and touchpoints are scattered. To address this issue, they built an all-scenario AI vacation intelligent agent, AI G.O, through Alibaba's Tongyi Qianwen large model and Lingyang's AgentOne platform. With a 24/7 instant response time of less than 1.5 seconds, they have entrusted a large number of repetitive and standardized tasks to the Agent, freeing up human resources for more personalized services.
Facing the increasing personalized and customized needs of Chinese enterprises, an important value of Lingyang's AgentOne platform is that it enables enterprises to independently build personalized AI workflows and corporate images, thereby quickly responding to business changes and enhancing their differentiated competitiveness. Fosun Tourism Group is a typical example of this capability. After initially verifying the service value of the Agent through cooperation with Lingyang, Fosun further developed an AI intelligent agent with a unique IP image based on the AgentOne platform and optimized it in combination with industry large models.
The entire process was completed and launched in about 90 days, which not only demonstrates the rapid delivery ability empowered by the platform but also balances brand personalization and system stability while pursuing speed. Ultimately, this intelligent agent effectively promoted the upgrade of customer experience and a substantial increase in the overall repurchase rate.
The same is true for Muyuan Foodstuff. The cooperation case between this industry giant and Lingyang deeply illustrates the irreplaceability of enterprise-level Agents in extremely complex business scenarios.
As a leading enterprise spanning 22 provinces, with 10 subsidiaries and covering nearly 80 regions, Muyuan Foodstuff's management system highly relies on data-driven decision-making. It holds nearly a hundred sales and management meetings every week, and more than a thousand people participate in the weekly management meeting simultaneously. The basis for each decision-making is a data analysis PPT of over 100 pages, covering all aspects of business views such as sales progress, profit achievement, category performance, and customer analysis.
Behind the data is a team of dozens of analysts constantly converting massive amounts of data into decision-making basis. However, as the business scale continues to expand, the traditional approach can no longer support the goal of refined operation, which aims to "serve every subsidiary and every regional manager." Slow data response, long report generation cycles, and inconsistent analysis standards have become the key bottlenecks restricting management efficiency and strategic implementation.
Lingyang's "Super Data Analyst," Agent Xiao Q, is the key to solving this pain point. This Agent algorithmizes and models the enterprise's mature business analysis framework and expert experience, constructing a dual-engine architecture of "Report Agent + Question Data Agent." After the system was launched, the analysis reports that previously took several days to generate manually can now be automatically generated within 30 minutes, precisely following the enterprise's unique management logic and analysis path.
The existence of Agent Xiao Q has not only helped Muyuan Foodstuff improve sales management efficiency by 80% and reduce ineffective meeting disputes by 50% but also achieved 90% coverage of self-service queries in business scenarios. By using AI to dig out customer risks and business opportunities in real-time, Muyuan Foodstuff has established a digital closed-loop from goal setting to execution tracking, driving the enterprise to achieve a leap from "uncontrollable experience-driven" to "predictable growth."
It can be seen that Lingyang's strategy is not to rush to win benchmark customers in a single field but to find and verify universal and implementable enterprise needs across various industries, refine vertical scenarios, and reduce complexity from commonalities, making AgentOne the "first stop" for AI applications.
03. What is still lacking for an Agent to "think"?
The third significant value of Lingyang comes from the strong ecosystem of Alibaba.
As a full-stack AI service provider, all B2B solutions on Alibaba Cloud are not isolated. They can be seamlessly integrated with other products in the ecosystem to jointly create end-to-end closed-loop solutions, greatly reducing the complexity for enterprises to deploy a single intelligent product.
Moreover, there is a wealth of data and valuable scenarios in Alibaba's business segments worth exploring, which have inspired Lingyang. The cooperation between Lingyang and Dianxiaomi is a good example.
Dianxiaomi is designed to meet the diverse business needs in the e-commerce scenario, especially focusing on the pre-sales to after-sales stages. Lingyang extracted the relevant capabilities related to Dianxiaomi's scenario and packaged them into an overall solution, enabling customers to leverage Alibaba Cloud's computing power. After the integration of capabilities, Lingyang, Dianxiaomi, and Hisense collaborated.
The combination of the advantages of multiple parties has built an unbreakable ecosystem.
The same applies to the Tmall New Product Innovation Center (TMIC). Due to Taotian's long-term accumulation of a large amount of data on clothing trends and its environment with a general large model and data privacy protection, it is very suitable for building Agents. Lingyang cooperated with TMIC and, based on the rich data of Taotian's e-commerce merchants, launched a new product innovation Agent on the AgentOne platform to help clothing brands keep up with fashion trends, quickly launch seasonal styles, and get them on the shelves rapidly.
Lingyang can better address the problem of data silos in enterprises. It is not only because of its own product and technical capabilities but also because of the openness and integration of the ecosystem, which gives Lingyang AgentOne a unique ecological position.
This also explains why Lingyang has an edge in solving the problem of enterprise data silos. It is not an isolated platform but an intelligent hub that can mobilize all relevant resources.
Take the