LynxAI: Let the cognitive equal rights brought by AI break through the bottleneck of product innovation.

#AI Applications# #Cross-border E-commerce#

Among the 38 species of felids in the world, lynxes are a unique genus. They belong to the Felinae subfamily (small cats), the same as domestic cats and Pallas's cats, but possess the hunting abilities of large felids. They are well - rounded in skills and highly resilient, making them natural hunting experts in the forest.

The startup "LynxAI" chose the English name of the lynx as its company name, aiming to become a small yet agile "predator" in the field of AI applications. "Today, AI technology enables a small team like ours to develop powerful AI applications," said Ding Mingming, the co - founder and head of product success at LynxAI, comparing it to traditional IT companies in the past that required thousands of R & D personnel.

Different from the high - profile and rapid advancement of large models, AI applications are quietly integrating into more and more vertical fields. In the relatively competitive "AI + cross - border e - commerce" track, LynxAI focuses on the more challenging product R & D segment. Based on its self - developed product innovation model and Agentic AI framework, it has launched a product innovation model and application platform positioned as an "AI product manager," providing systematic solutions such as market opportunity insights, product innovation portfolio optimization, and market operation strategies for cross - border e - commerce enterprises.

Zhang Keyi, the founder of LynxAI, said, "We didn't engage in AI sales or AI customer service because we want to target the core of value creation - product innovation. This will be the most challenging path, but in the long run, it's the truly viable one."

Zhang Keyi, the founder of LynxAI, has previously served as the COO of ZUO, the Chief Strategy Officer of Anker Innovations, and the General Manager of Strategic Execution at Fosun Group.

The differentiated product positioning and systematic solutions have also attracted more attention to LynxAI. In the month when the news of its angel - round financing was disclosed, more than 40 investment institutions expressed their intention to contact. This both excited and frustrated Zhang Keyi. "Since the registered address is my home address, some investors even came directly to knock on my door."

Good projects are never short of market demand. The influx of customer needs has also accelerated the refinement and implementation of the products. LynxAI plans to officially launch its first product, the Global Market Insight Expert 'Jaxx,' in May.

1. Shorten the R & D cycle and improve the success rate of new products

In the past decade, cross - border e - commerce sellers generally searched for the same products on 1688 based on the best - selling data on platforms and then mass - distributed them through multiple stores. However, with changes in the market environment such as the cancellation of the $800 small - value exemption clause, this approach has become increasingly unsustainable. Some large cross - border sellers with annual revenues of hundreds of millions of yuan have begun to shift their focus to product refinement, new product development, and brand incubation, hoping to break through their development bottlenecks and escape the intensifying price competition.

"All consumer - oriented companies dream of being the first to sense the changing trends in consumer demand and make preparations in advance, rather than following the trend passively," said Ding Mingming, taking the market opportunity insight of the car - carrier cat cage category as an example. "During the Spring Festival this year, the topic of 'the first batch of cats going back to the countryside for the Spring Festival' became popular. An excellent product manager would realize that the pet travel scenario might be booming, which reflects a change in the relationship between humans and pets. Then he would check the sales data of the 'car - carrier cat cage' category on e - commerce platforms. Finding a rapid growth, he could develop new products such as anti - motion - sickness features and automatic water dispensers for this scenario. Finally, he could verify his judgment through consumer feedback on social media."

Ding Mingming, the co - founder and head of product success at LynxAI, has previously served as a digital business partner at Oookini, a co - founder at Guokr, and the head of mobile business at Tudou.com.

It can be seen that this new product development model largely depends on the personal ability of product managers. "A company may spend years cultivating an excellent product manager, but they may leave at any time. It's even more difficult for small and medium - sized enterprises to find such talents," said Zhang Keyi. Even for experienced product managers, it's difficult to collect social media and consumer data from overseas markets, which easily limits the accuracy of market and consumer insights. If a company chooses a third - party consulting or marketing company like Nielsen or GFK for market research, the budget for a single project usually ranges from hundreds of thousands to millions of yuan, which is a significant cost for enterprises.

The low success rate of new products has always been a common problem in the cross - border e - commerce industry. According to data released by TikTok, if products with an average daily sales volume of more than 30 orders in the past 30 days and a growth rate of more than 30% in the next 30 days are defined as best - sellers, the proportion of best - sellers in the overseas market is as low as less than 0.6%.

"So we're thinking about whether we can use AI to implement the best practices of product managers in the industry, shorten the R & D cycle of new products, and thus improve the company's success rate of new products within a certain period," Zhang Keyi introduced. To address this common pain point, LynxAI's products rely on the self - developed MOE mixed expert model and Agentic AI framework. Through technical means such as algorithms and intelligent rules, they break down the successful practices verified by the market in different consumer goods industries into AI workflows capable of handling complex tasks.

After an enterprise proposes a task, the Global Market Insight Expert 'Jaxx' will intelligently collect external data in real - time, integrate the enterprise's internal data and rules, conduct comprehensive reasoning, and finally output a visual analysis report for the enterprise's decision - making reference. Zhang Keyi revealed that according to the actual test data of existing customers, LynxAI can increase the response speed of complex business analysis to the minute - level, which is equivalent to completing the market research and consumer insight work that a product manager usually takes two weeks to finish in just two hours, helping product managers improve work efficiency and reduce trial - and - error costs.

The reports generated by LynxAI will provide multiple structured market opportunity insights to assist enterprises in decision - making. As the application gets to know the enterprise better, it can continuously optimize its learning and output capabilities in the future, giving priority to recommending targeted solutions that are more suitable for the enterprise's financial situation, organizational characteristics, and supply - chain capabilities, and even becoming an assistant for providing R & D inspiration.

Currently, the challenge for LynxAI is whether it can quickly standardize the configurable modules of its relatively non - standard products through data accumulation and model optimization, so that customers can use them just by connecting to their own data systems. "This is also our most core goal at this stage," Zhang Keyi said.

2. An AI employee capable of autonomous delivery and continuous learning

In the view of LynxAI, the ability to self - learn and strengthen based on data and user feedback, and continuously improve the delivery quality, is the advantage of Agentic AI over traditional data tools. Zhang Keyi explained, "Traditional digital tools focus on process transparency, and people's work delivery results are hosted on digital tools. In contrast, Agentic AI focuses on understanding complex tasks, and each Agent can autonomously complete task delivery. To achieve this, we integrate the product with industry rules and data to avoid hallucinations, use non - linear feature extraction methods to mine implicit associations, and ensure the bidirectional traceability of reasoning conclusions and processes."

Taking the Product Market Insight Expert Jaxx, which LynxAI is about to launch, as an example, when analyzing the opportunities of a specific niche category, Jaxx abstracts the customer's demand for market opportunity insights into structured and open - ended questions. "For example, the penetration rate of large long - haired cats in China is less than 10%, while in North America, it is as high as 40%. Discovering the market opportunity for large - sized cat litter boxes in North America is called structured insight," Ding Mingming explained.

After determining the question type, Jaxx will intelligently decompose the market map based on real - world market data, find the blank price bands, conduct opportunity ratings, quantitative and qualitative analyses of consumer insights, and integrate the customer's supply - chain practices to rank the customer's new product opportunities. For scenarios such as annual strategic business meetings, Jaxx can plan new categories and analyze niche categories, helping cross - border brands formulate strategies and verify whether the current strategies match the market trends.

"I have more than a decade of experience in product engineering and advocate for product interaction and user perception that 'don't make me think,'" Ding Mingming introduced. To enable Jaxx to automatically complete the above analysis process, the user's operation process must be simplified. The product goal is that a product manager only needs to input an Asin List (Amazon product ID), and Jaxx will conduct a four - dimensional analysis and reasoning process of market insight, consumer insight, supply - chain insight, and technology insight with this group of products as the origin, until it provides the enterprise with a "ranking of actionable opportunities."

"For example, when analyzing user reviews, we will tell Jaxx that one - star reviews may indicate customer dissatisfaction, five - star reviews may be fake, and long three - star and four - star reviews often better represent the customer's real experience and should be given more weight." The Jaxx expert small model will analyze the data based on a large number of such detailed domain rules, using multiple knowledge models and proprietary algorithms to find the implicit associations behind and dig out potential market opportunities.

A schematic of the interface of the Global Market Insight Expert 'Jaxx,' the first product that LynxAI will officially launch in May.

The cognitive framework of the domain is the key for AI applications to think and analyze problems like real product managers. Liu Tong, the Chief AI Scientist at LynxAI, has an academic background in applied mathematics, marketing, computational advertising, and human - computer interaction, and has completed post - doctoral research in interdisciplinary fields. He is responsible for designing industry intelligent rules and algorithms at LynxAI. "Currently, cross - border e - commerce customers generally feed VOC (Voice of Customers) or social media data directly into large models, which leads to poor results and high computing costs. If only the enterprise's knowledge - base data is used, as the data volume increases, the quality of the retrieved content will also rapidly decline."

Liu Tong, the Chief AI Scientist at LynxAI, is a post - doctoral fellow at Zhejiang Lab, a doctor from Wuhan University in the fields of A1, human - computer interaction, and user behavior, and specializes in artificial intelligence, computational advertising, and consumer behavior theory.

Therefore, after intelligently obtaining consumer data, LynxAI first processes and cleans it using a large number of detailed industry rule sets, and then conducts multiple rounds of recursive reasoning using self - developed models and algorithms until it provides accurate consumer insights. As the analysis dimensions expand and the insight perspectives change, Liu Tong's team needs to continuously optimize the algorithms, and more logical branches and algorithms need to be incorporated into this self - developed model.

To pursue the accuracy of content and the sense of business insight when using the product, LynxAI has separated the content engineering and Agent teams from the AI technology engineering architecture. Teams with industry or academic backgrounds in consumer market research, computational advertising, and consumer behavior research are responsible for optimizing the content output quality of each Agent, speaking in the tone of a real product manager rather than the programmer - style writing of general large models. "In the past, we focused on software engineering quality and product engineering quality. In the AI era, the content engineering quality of professional AI products should be given new importance," Ding Mingming said.

It can be said that LynxAI hopes that the differentiation of its products lies in creating an AI employee capable of autonomous delivery and continuous growth by integrating the team's interdisciplinary knowledge in the industry and AI technical capabilities, rather than just providing consulting services packaged in AI or traditional SaaS products assembled with large language models.

3. More predictable than humans and more understanding of enterprises' businesses than platforms

After completing the acceptance of the product prototype in October last year and ensuring that the product could integrate with the customers' business rules, LynxAI conducted a one - week PMF (Product Market Fit) verification. "To be honest, the PMF results far exceeded our expectations," Zhang Keyi revealed. Currently, LynxAI has reached cooperation agreements with more than a dozen cross - border enterprises, with scales ranging from hundreds of millions to tens of billions of yuan, covering multiple industries such as household items, cleaning equipment, hair - styling tools, pet supplies, and outdoor backpacks.

The reason is that the LynxAI team has both technical capabilities and e - commerce industry know - how. Their cross - domain composite background makes these enterprises more willing to share their business data. On the other hand, the product positioning largely addresses two major pain points of cross - border e - commerce enterprises: the short - term pursuit of efficiency and the long - term precipitation of enterprise knowledge.

In the short term, the challenge for cross - border e - commerce enterprises lies in growth anxiety and the resulting efficiency competition. "Cross - border brands are generally accustomed to using technology to drive efficiency improvement and gain greater competitiveness. Therefore, from the very beginning, they have been more open, even eager, towards AI technology," said Yang Binglong, the founder of Huajun Chuhai. Compared with traditional SaaS tools, AI tools have a simpler interaction method, only requiring dialogue and task - setting, so the customer training and education costs are also lower. "In addition to the marketing scenarios most involved in AI applications, many cross - border companies also hope that AI can help discover global market opportunities and assist enterprises in product definition."

In the long - term development, Zhang Keyi believes that AI applications can make the successful practices and implicit resources in product managers' minds explicit through AI tools, helping enterprises precipitate their knowledge bases and reducing the operational instability caused by personnel turnover.

Currently, LynxAI mainly reaches customers through the resources of its founding team and angel - round investors in the cross - border e - commerce industry. At the same time, it cooperates with various industry associations and platforms and uses industry conferences to expand potential customers.

The company's future growth path is mainly divided into two directions: one is to cooperate deeply with large - scale leading cross - border brands to continuously refine industry best practices and sample cases; the other is to upgrade the existing products to a subscription - based platform - standard service to directly empower small and medium - sized enterprises in product innovation.

"Ideally, the higher the standardization level of the platform - based product, the more customers there will be, and the lower the sales and delivery costs will be," Zhang Keyi said. Next, LynxAI also plans to cooperate with large - scale AI application platforms to be embedded in enterprise IM systems. "In this way, it will be more like a real AI colleague."

In terms of team planning, LynxAI plans to focus on recruiting more algorithm engineers with knowledge of the cross - border e - commerce industry and consumer markets, "that is, more interdisciplinary talents like Dr. Liu Tong." Ding Mingming said that one of the reasons LynxAI chose to come to Hangzhou was to be closer to the professional talents from Zhejiang University and several universities in Xiasha. In addition to interdisciplinary talents, LynxAI also hopes to attract more investors who "understand both AI application