Original · Unique Research · 2025-11-09
Editorial note: Company revenue and market observations are retained as historical statements from the source, not independently verified current results. The narrative describes Japanese project cycles reaching a year, while Q4 specifies a sales cycle of at least a year; both formulations are preserved. In Q10, the original wording suggests serving early customers for free or even paying to do so, but does not explain the payment arrangement. The source’s descriptions of regional markets reflect the interviewee’s experience rather than universal market rules. The omitted ordinary conference portrait is labeled CAAIEC: China (Guangxi)–ASEAN Artificial Intelligence Enterprise Conference and carries no additional business facts.
In this AI era filled with CEOs born in the 2000s and tech prodigies born in the 1990s, one entrepreneur born in the latter half of the 1980s is quietly charting a different path to globalization. He is not the protagonist of a Silicon Valley myth in which a startup becomes an overnight sensation, nor is he interested in chasing financing myths. Instead, over six or seven years he has refined a system of AI products designed for practical scenarios and gradually deployed around the world. His name is Liu Siping, founder of 94AI.
This conversation is not a get-rich-quick guide to “making a fortune through an AI startup.” It is more like a calm retrospective by a middle-aged entrepreneur on what it means to “go global” amid rapid change. He is not sentimental, yet every sentence rings true.
I. Why Go Overseas? Because You Cannot Keep Tilling the Same Plot with the Same Hoe Forever
Founded in 2018, 94AI decided to go overseas in 2021. At the time, it already had stable domestic customers, a clear product model, and revenue of roughly RMB 50 million. This was not a case of “running away because the domestic business had become untenable,” but of “seeing opportunities overseas after validating the business at home.”
Liu Siping said he saw two trends at the time. First, the company’s AI voice products could already be deployed reliably in finance, e-commerce, and education. Second, digital demand in emerging markets such as Southeast Asia resembled China’s demand 3 to 5 years earlier: “Our experience and technology could be applied there directly.”
This is an important point of departure: going overseas should not mean “betting on the next market,” but “extending the boundaries of capabilities you already possess.” You cannot expect to arrive abroad and suddenly become a different person. But you can expect your capabilities to be revalued in a different ecosystem.
II. The First Lesson of Going Global: Southeast Asia Is a Familiar Trap; Europe and the United States Are Distant Opportunities
Many Chinese AI entrepreneurs choose Southeast Asia as their first destination for simple reasons: the culture feels close, the Chinese diaspora is large, and the market seems familiar. 94AI was no exception. But Liu Siping later admitted, “A place that looks easy is not necessarily a place where you can go far.”
Southeast Asia is indeed a good place to validate a model and refine a product, but competition there is as brutal as it is in China. European and U.S. markets have high entry barriers and wide cultural differences, yet they may offer greater margins and stronger pricing power. Japan is an archetypal “slow market”: a project cycle can last as long as a year and customers are extremely conservative, but once a partnership reaches the table, customer loyalty is also high.
This implies a philosophical question for entrepreneurs: do you take the easy road first, knowing that fierce competition awaits later, or take the difficult road first, with the possibility that it will widen as you proceed?
Different choices lead to completely different organizational and product designs.
III. The Misunderstood “Translation”: It Is Not Language Translation, but Cultural Localization
During its early overseas expansion, 94AI made a classic mistake: assuming that Chinese-language materials could simply be “translated” into English and put to use. The real challenge, however, was not language. It was translating culture, communication styles, and industry experience.
The “playbook” of China’s financial industry will not necessarily work in Brazil or Indonesia. The industry know-how you accumulated at home can sometimes become a “language barrier” when you stand before overseas customers.
True localization does not mean hiring a translator; it means “hiring a local.” 94AI later recruited part-time and even full-time employees in its target markets specifically to help refine its AI training materials and dialogue logic. This form of “human, embedded” localization looks resource-intensive, but it is precisely what is required to make AI genuinely “human-like.”
IV. From Small Models to Large Models, AI Moves from Tool to Teammate
Liu Siping offered a vivid analogy: earlier AI was like an “elementary-school student.” It could do repetitive work, but conversations had to be tightly constrained. Today’s large models, however, can function like “graduate students,” participating in deeper communication and even decision-making.
In the past, achieving reliable results required “narrowing the task scope.” Today, it means “deepening the conversation.” This has transformed the voice Agent from a “robot that can talk” into a business partner capable of completing tasks and generating conversions.
In the past, AI supported the business. Now AI leads certain parts of the business process. This is not merely an upgrade in speech recognition; it is an evolution in business logic.
V. A More Middle-Aged Approach to Entrepreneurship
This era is extraordinarily friendly to young people: financing, buzz, and media attention are concentrated almost entirely on 20-year-old entrepreneurs. Yet this conversation reveals a completely different entrepreneurial approach—moving a little more slowly and steadily, with operating capability and a long-term mindset.
Liu Siping’s experience is not about “racing against time to seize speed.” It is about “using China as a training ground, standardization as the spear, and localization as the shield” to open overseas markets one step at a time. He did not bet on producing a breakout product or use financing to chase scale. Instead, from the perspective of a middle-aged entrepreneur, he asked “how experience can be exchanged for efficiency.”
When Wu Wei asked whether, if he could start again, he would enter European and U.S. markets first, Liu Siping said yes—but he also emphasized that “it is difficult to optimize risk, return, and liquidity all at once.”
VI. Entrepreneurship Is Not Standing in a Tailwind and Waiting for It to Blow, but Finding a Road You Can Keep Following
The conversation contained no grand narrative or impassioned slogans, yet it landed like a stone in the still pool within many AI entrepreneurs’ minds.
Perhaps you are not the youngest, the most cutting-edge, or the best storyteller. But if you design your product for global markets from the beginning, first find customers in China with whom you can validate it, and then use standardized products and localized services to break into overseas markets, it does not matter if you move a little more slowly. What matters is that you go far.
The future of AI may not belong to those who run fastest, but to those who know most clearly where they are going.
VII. Selected Interview Q&A
About 94AI and Its Decision to Go Global
Q1: Under what circumstances did 94AI decide to begin expanding overseas? How was the company’s domestic business developing at the time?
Liu Siping: 94AI was founded in 2018 and began experimenting with overseas expansion in the second half of 2021. At the time, the company already employed dozens of people in China, and its revenue for 2021 reached approximately RMB 50 million. The decision to go overseas was driven mainly by two factors. First, the company’s product technology, especially its AI applications in finance, education, and e-commerce, had been thoroughly validated in China and was performing well. Second, founder Liu Siping observed that the e-commerce and financial markets in Southeast Asia and elsewhere were developing rapidly and resembled China 3 to 5 years earlier, leading him to predict substantial local demand for AI.
Q2: What was the initial catalyst for the overseas expansion? Was it proactive market development, or did customers lead the way?
Liu Siping: The initial catalyst came from customers. Some e-commerce and financial customers that were already working with 94AI in China found that they also needed AI voice services for customer service, telemarketing, and other scenarios when they expanded overseas. Because a foundation for cooperation already existed, they approached 94AI hoping to continue the relationship abroad. This made 94AI’s overseas expansion relatively smooth: the product foundation and application scenarios were highly similar, so only language and localization adaptations were required.
Experience and Insights from Different Markets
Q3: While developing overseas markets, what do you see as the greatest difference between overseas and Chinese markets when acquiring early customers and conducting POC projects?
Liu Siping: The greatest difference lies in how customers accept new suppliers and in their business practices. In China, early market development is relatively easy. If you possess fairly new technology, many customers are willing to offer opportunities for free testing and trials, or POC projects; when there is no charge, finding seed customers is especially easy. Overseas, however, the situation is exactly the opposite: even free POC opportunities are almost impossible to secure. Overseas customers approach a POC very cautiously, impose high preliminary requirements on suppliers, and are accustomed to paying for the POC. The domestic market can therefore be described as “easy first, difficult later”—easy to enter, but fiercely competitive afterward—while overseas markets are “difficult first, easy later.”
Q4: You have visited dozens of countries. Could you share your observations about the differences among the distinct markets of Southeast Asia, Latin America, and Japan?
Liu Siping:
Southeast Asia: This is the region most similar to the Chinese market because many ethnic Chinese entrepreneurs and Chinese companies operate there. That gives Chinese enterprises a relatively low entry barrier, but the disadvantage is also similar to China: subsequent market competition can become intense.
Latin America: This market lies between Southeast Asia and Europe and the United States, and it has strong potential. Per-capita GDP and income levels in Latin America are higher than in Southeast Asia, giving enterprise customers greater ability to pay. Correspondingly, however, the market is harder to enter and cultural differences are greater. Relatively few Chinese companies expanding overseas operate there, so acquiring early customers is more difficult; once a company establishes itself, however, its competitive barriers are also higher.
Japan: This is an extremely conservative market, especially in enterprise business. Entry barriers among enterprise customers are exceptionally high, even higher than in the United States. A new company needs a sales cycle of at least a year to sign a Japanese customer. Japanese customers also tend to work with local channel partners or agents rather than foreign companies directly, so local channels are the key strategy for market entry.
Q5: What is your view of the U.S. market for AI startups seeking to enter it?
Liu Siping: Whether for consumer or enterprise AI startups, the United States is an excellent market with customers that have strong purchasing power. But the market is characterized by “you either win everything or lose everything,” with little middle ground. Many products must adopt a platform model; the companies that ultimately break through grow very large and raise substantial financing, while there are relatively few mid-sized players. For Chinese entrepreneurs, directly entering the market for medium and large customers in Europe and the United States is extremely difficult, but starting with small and medium-sized customers is considerably easier.
Products, Technology, and Challenges
Q6: What was the biggest “pitfall” you encountered in the early stages of taking your product overseas, and how did you solve it?
Liu Siping: The biggest “pitfall” was underestimating the depth of “localization.” At first, the team believed it could simply translate into local languages the dialogue content and industry knowledge bases already validated in China. That approach presented few problems when serving Chinese customers expanding overseas, because everyone shared similar cultural and business logic. But when we faced truly local customers, we discovered that the translated content was completely inconsistent with local communication habits and culture and sounded highly “inauthentic.”
To solve the problem, the company later began hiring local part-time staff and gradually developed full-time local teams. Those employees optimized and calibrated the content, producing a tremendous improvement in results.
Q7: From 2021 to the present, what core changes has the development of large-model technology brought to your AI voice products?
Liu Siping: Technological progress has produced a qualitative leap. Liu Siping describes earlier AI as an “elementary-school student” with limited conversational ability; the scope of a scenario had to be strictly constrained to ensure good results. By contrast, today’s large-model-based AI is like a “university student” or even a “graduate student.” Its depth of conversation, fluency, and human likeness are so advanced that users already find it difficult to tell they are speaking with AI. This has not only improved business value in areas such as conversion rates and customer satisfaction; more importantly, the product has evolved from a simple “conversational robot” into a genuine “voice Agent.” It can converse, but it can also perform analysis before a conversation, execute tasks during it, and produce a summary afterward, allowing it to participate much more deeply in customers’ business processes.
Q8: In your view, where has the central challenge in AI technology shifted from the technology itself?
Liu Siping: The central challenge has shifted from “building an AI with strong conversational ability” to “deeply integrating that powerful capability with a specific business scenario to create real results and value.” Liu Siping believes that building a technically impressive conversational AI alone is no longer difficult. What is truly difficult is understanding an industry’s pain points, seamlessly integrating AI into its business processes, and maximizing both efficiency and outcomes. That is now the most critical task.
Reflections and Advice
Q9: Looking back, if you could choose your route overseas again, how would your thinking differ?
Liu Siping: Although he acknowledges that entering Southeast Asia by “following customers” was the correct decision at the time, Liu Siping says that from today’s perspective he would consider another route: entering developed European and U.S. markets directly. The playbooks for the two routes are completely different.
The Southeast Asia route: Customers are primarily medium and large enterprises, and the service model needs to be fairly “heavy” to satisfy their requirements. The advantage is a relatively high probability of success.
The Europe and United States route: A company can start with small and medium-sized customers that have strong purchasing power and build a more standardized, platform-based product. Although this road is difficult, success offers much greater potential. Choosing European and U.S. markets also means establishing a long-term local presence and accepting greater risk.
Q10: What concrete advice do you have for AI entrepreneurs who want to Go Global from Day one?
Liu Siping recommends a route he considers highly effective, especially for entrepreneurs in enterprise markets:
Design globally, validate locally: Build a global perspective and architecture into the product from the initial design stage.
Refine the product in China first: Do not begin by looking for customers overseas, because both researching demand and securing POC opportunities are extremely difficult. The right approach is first to find comparable enterprise customers in China, serve them for free—or even pay to do so, refine the product until it is mature, and develop a relatively complete standardized offering.
Go overseas after the product matures: Expand into overseas markets—through whatever channel—with a mature product that has already been validated, and you will face far less resistance. Without a mature product or a deep understanding of local customer needs, it is very difficult to find the right point of entry.