Original · Unique Research · 2025-11-09
Editorial note: This historical article preserves the source’s reporting and the interviewee’s claims, not independently verified current results. The narrative states a CPL reduction of more than 30%, while Q4 gives “at least 10%–30%”; the source does not reconcile these formulations. The nearly 50% dealership penetration claim has no specified geographic scope or denominator. The opening comparison with Tesla’s European sales does not identify the automaker, reporting period or vehicle category. TikTok usage and Southeast Asian demographic figures remain attributed to Shen Yang. The amount in Q4 has no specified currency. Qiyu Digital Technology is a conservative rendering of the Chinese company name. The omitted ordinary conference portrait carries the label CAAIEC: China (Guangxi)–ASEAN Artificial Intelligence Enterprise Conference, but no additional business facts.
When people discuss AI technology, many talk about Sora2 and Veo 3.1’s video-generation capabilities, ChatGPT and DeepSeek’s advances in reasoning, and how large models are making the world more intelligent. But the AI that is genuinely being deployed is not in a California conference room. It may instead be in the 4S dealership downstairs from your home, in the backend of a Douyin account where a user casually leaves a comment after watching a short video, or in the sales chain taking a Chinese new-energy vehicle overseas—where European sales have surpassed Tesla’s—quietly completing a real conversion.
In my conversation with Shen Yang, founder of Qiyu Digital Technology, what struck me was not the “wow factor” of the AI technology itself, but how it had quietly embedded itself precisely where a traditional industry needed it most. Like water, it was silent yet powerful, nourishing everything while wearing through stone.
AI and Sales Are No Longer “Two Separate Things”
The pain points of traditional automobile sales are obvious: declining dealership traffic, high offline conversion costs, difficulty producing marketing content, and unfamiliarity with new-media tactics. For many automotive brands and 4S dealerships, transforming into a “content-commerce enterprise” is almost an impossible mission. They neither have people who understand new media nor the ability to retain those people.
Qiyu Digital Technology, the company founded by Shen Yang, has made that mission possible with AI agents. What it does is not “teach enterprises to use AI,” but “have AI do the work for enterprises.”
Not shooting short videos? AI shoots them for you.
Do not know how to edit? AI completes the edit automatically.
No one available to receive customers during a livestream? AI welcomes them in the comments, direct messages, and livestream itself, even leading the conversation and guiding users to leave their contact details.
Customers may not even know they are chatting with AI.
The key is that the results really can be better than those delivered by people: faster response times, a lower cost per lead—CPL falls by more than 30%—and more efficient content iteration. Most importantly, people do not have to teach AI how to do the work; instead, AI shows the enterprise how to transform.
“An Agent Is Not a Tool; It Is an Employee Who Gets Things Done for You”
Shen Yang repeatedly emphasized one term: “closed loop.” Many AI applications fail because they “solve only half the problem.” The value of an Agent is that, within a complete value-generating scenario, it finishes every part of the job. From asset generation to content distribution, from user interaction to lead management, and then to the final sales conversion, the Agent’s real power emerges only when AI assumes responsibility for the entire chain.
This also explains why the company has achieved a penetration rate of nearly 50% among 4S dealerships. It is not that the businesses have come to understand AI, but that AI has become an employee to whom they can entrust outcomes. This model—”doing the work for you and owning the KPI for you”—is the kind of AI agent that has genuine commercialization prospects.
The Underlying Logic of Deployment: Not Showing Off Technology, but High-Frequency Demand + Data Accumulation
One interesting detail is that Qiyu Digital Technology did not begin building Agents only after GPT became popular. As early as 2021, it was already accumulating real-world data from livestreams, direct messages, comments, and other interactions. Those data later became the “nutrients” used to train large models and build Agents. Whether an AI can communicate like a salesperson does not really depend on which model it uses; it depends on whether it has heard, observed, and participated in enough real sales conversations.
AI has never been a “superpower that fell from the sky”; it is something that “grows out of a scenario.” Nor is an Agent “a miniature version of artificial general intelligence”; it is “a virtual employee that can work for you at the point where your pain is greatest.”
Taking the Chinese Model Overseas Is a Dimensional Advantage, Not a Simple Transplant
International expansion is another thread. In Shen Yang’s view, China’s AI+ content-commerce playbook has a “dimensional advantage” in Southeast Asia. China’s content ecosystem and AI capabilities are already far ahead, while the structure of the ASEAN market—a young population, strong attachment to social media, and genuine demand for new-energy mobility—makes the model a natural fit.
As new-energy vehicle brands expand overseas together, they also need new ways to acquire customers. Qiyu Digital Technology follows automakers into overseas markets, using a model already validated at home to deploy AI-agent sales systems quickly in local markets. There is no need to reinvent the technology. The company can replicate the “methodology” proven in China, then keep optimizing it through data and feedback, moving from 60 points to 80 and then 90.
This is what it truly means for AI to go global. It is not exporting an algorithm or a model; it is embedding in a market an entire capability for “using AI to solve business problems.”
Will the Advertising Industry of the New Era Be Made Up of Technology-Led Integrated Marketing Companies?
When we discussed whether traditional advertising agencies would be replaced, Shen Yang’s answer was “coexistence.” In the short term, AI will indeed struggle to replace premium advertising content with high artistic value and rich emotional expression. But AI is best suited to the enormous long tail of video assets and marketing content.
That produces an intriguing picture: the integrated marketing of the future will no longer mean “dreaming up an idea on instinct.” It will mean “letting AI handle mass production + data-driven optimization, with creativity supplying only the finishing touch.” Such a new type of company might even employ both an AI engineering team and an advertising creative director, reshaping the marketing-services industry through a combination of “creativity + technology.”
Shen Yang said they have not reached that point yet. For now, it is more important to concentrate on taking their capability for “standardized content at scale” to its limit, because that is where the marginal effects are greatest and the commercial value clearest.
But what about the future? Who knows whether a fully AI-driven advertising group might emerge from a company like theirs?
AI’s Real Breakthrough Point Is Not Silicon Valley, but the Scenario
We tend to assume that AI’s breakthrough point lies in “more powerful models.” In reality, breakthroughs often occur where needs are more painful and scenarios more frequent.
China’s 4S dealerships, short-video platforms, and livestream studios form fertile soil in which AI can most readily grow. Enterprises that never understood AI are being quietly transformed by “Agents that get things done.” These seemingly unremarkable “deployment projects” are in fact building the most solid foundation of the AI industry.
Not everyone can build GPT, but everyone can find a real problem and solve it with AI.
That is the force most worthy of attention in the AI era.
Selected Interview Q&A
Q1: What is Qiyu Digital Technology’s core business, and which pain points does it address in the automotive industry?
Shen Yang: Our core business is to use “AI + new media” technology to build an intelligent Agent in a scenario centered on automobile sales, helping automotive brands and 4S dealerships conduct online marketing and customer acquisition.
We primarily address the traditional automotive industry’s central pain points as it transitions to new media:
Traffic depletion: Organic traffic at offline dealerships is shrinking sharply, while online traffic—especially traffic on new-media platforms—is surging.
Difficulty transforming: Automakers and 4S dealerships generally lack teams and experience in new-media operations. Even when they recruit the right people, differences in compensation systems and working models make those employees difficult to retain.
Low efficiency: Traditional methods of producing content and receiving customers cannot meet the 24/7, high-efficiency, standardized operating requirements of new-media platforms.
Q2: Why does Qiyu Digital Technology regard Southeast Asia as an important destination for overseas expansion?
Shen Yang: We believe Southeast Asia is a market with enormous growth potential, primarily for the following reasons:
A vast and highly active user base: Taking TikTok as an example, Southeast Asia has as many as 450 million monthly active users, who spend about 4 hours per day on the platform on average. This creates fertile ground for new-media marketing.
A young population: 50% of Southeast Asia’s population is under 30, and these people are highly receptive to new media, new content, and new technologies.
The rise of Chinese new-energy vehicles: Thanks to their strong value for money, Chinese new-energy vehicles have an almost overwhelming presence in Southeast Asia. These automakers are already our customers in China. Their overseas expansion naturally creates marketing demand, and we are moving with that trend.
A mature business model: The “new media + e-commerce + AI” model has already been validated in China and can be applied with a “dimensional advantage” in Southeast Asian markets following a similar development path.
Q3: How exactly does your AI Agent operate in content production? Can it really replace human labor completely?
Shen Yang: Our AI Agent can provide customers with fully managed content production. For example, customers do not need to shoot footage, write copy, edit video, or add music themselves. We can use AI to generate an entire AI-generated short video suitable for commercial paid distribution.
At present, the process still cannot replace human labor 100%. Our people mainly intervene in the final “review” stage. Like quality inspectors on a production line, they remove substandard “waste footage” and tag qualified content so that the model can continue improving. Our ultimate goal, however, is to have a large model handle even this review step.
Q4: Can AI-generated content match the realism and performance of work by professional advertising agencies? How great is the cost advantage?
Shen Yang: In terms of realism, thanks to this year’s explosive breakthroughs in AIGC technology, the commercial paid-distribution assets we generate are now largely indistinguishable from content made without AI and have reached a usable standard.
In terms of performance, our accumulation of large volumes of data gives us a deep understanding of user preferences, so our content may convert even better than content created by a dealership’s own marketing staff. We promise customers that using our service will reduce the cost per lead, or CPL, by at least 10%-30% compared with traditional methods.
In terms of cost, the advantage is orders of magnitude. A blockbuster-style commercial shot by a traditional advertising agency may cost hundreds of thousands (currency unspecified in the source), while our AI-generated content primarily incurs Token costs. The difference is enormous.
Q5: What role does the AI Agent play in customer communication and lead conversion, and how is it more effective than human customer-service representatives?
Shen Yang: When users interact in a customer’s livestream or beneath its short videos, our AI Agent plays the role of a “salesperson,” receiving and guiding them in real time, with the ultimate goal of obtaining the user’s mobile number as a lead.
It is more effective than a person in two main respects:
Response timeliness: Peak activity on new-media platforms often occurs late at night or in the early morning, outside normal working hours. AI can respond within seconds, around the clock, seven days a week, ensuring that no potential customer is missed.
Communication expertise: We train the large model on vast amounts of real sales-dialogue data, enabling it to understand customer intent deeply and guide customers through the best scripts and processes like a top-performing salesperson. Its conversion capability is consistently stable and professional, avoiding fluctuations in human skill and the effects of emotion.
Q6: How do you teach a large model to behave like a “top-performing salesperson” rather than a “mechanical customer-service agent”?
Shen Yang: Our core moat lies in data accumulation. From 2021 to 2023, before going All in on large models, we spent more than two years accumulating vast amounts of automotive-industry data: livestream and short-video interactions, direct messages, and audio and video streams. These authentic, vertical-scenario data became the “nutrients” with which we trained the large model. They enable our AI Agent not only to understand technical terms and customer “slang,” but also to learn the most effective sales communication strategies and guidance techniques.
Q7: What is Qiyu Digital Technology’s business model? How do customers pay for these AI services?
Shen Yang: We offer two flexible payment models:
Monthly subscription: Customers pay a fixed monthly service fee for a fully managed service.
Payment by results: We charge for the qualified sales leads ultimately delivered, on a cost-per-lead (CPL) basis. This model presents lower risk to customers and is more attractive.
Q8: What does the “fully managed” service you mentioned mean in practical terms for customers?
Shen Yang: “Fully managed” means the customer has very little to do. Customers simply pay and authorize us to connect their Douyin and other new-media accounts; after that, they essentially do not need to manage anything. From content generation, publishing, and interaction to customer reception and lead collection, our AI Agent automates the entire process. Customers need only wait for the sales leads we deliver.
Q9: How do you localize for linguistic and cultural differences in overseas markets?
Shen Yang: We believe that although markets differ in language, culture, and user habits, the methodology for optimization is universal. In China, we have already developed an effective process and methodology that can continuously improve conversion. When we apply that methodology overseas, the initial result may score only 50 or 60 points, but we have a clear path for rapidly optimizing it to 70, 80, or even higher by continuously accumulating localized data.
Q10: How do you view the future relationship between AI marketing companies and traditional advertising agencies: replacement or coexistence?
Shen Yang: I believe they will coexist. AI excels at solving the mass-production challenge for enormous volumes of standardized, mid- and long-tail content, dramatically improving efficiency in that part of the market. But in the short term, AI still cannot replace people in premium commercial productions that demand top-tier creativity, meticulous craftsmanship, and complex human emotion. A division of labor will therefore emerge: AI will handle “scale” and “efficiency,” while the best human creators focus on premium-priced “high-end work.”