JISHIYU: Embracing Technological Change, Serving Users with Dedication

The transformation of industries through upgrades has become an undeniable trend at both the national policy and market selection levels. Within this trajectory, a cohort of forward-thinking enterprises has risen to prominence. In the expansive market, one finds not only headline-grabbing corporate giants but also numerous tech-innovative startups with a global vision.

The EqualOcean interview series focuses on startups that exhibit technological and product innovation, global competitiveness, and a worldwide presence in advanced manufacturing, new energy, consumer technology, and digital sectors. Through close observation of their technological product development processes and the global perspectives and visions of their core management, the series showcases the new appearance and opportunities of China's sci-tech entrepreneurship. In this edition, EqualOcean engages in a dialogue with Bai Tao, the founder of JISHIYU, an AI-powered customer service provider.

As an angel investor in JISHIYU, Ian Goh, Founder of 01VC, commented, "The JISHIYU team has accumulated years of expertise in the intelligent customer service field. Combining in-house AI products with integrated large language models, the team has developed an all-encompassing intelligent customer service platform tailored to the cross-border e-commerce industry. This platform efficiently aids cross-border e-commerce sellers in enhancing efficiency and elevating brand and user experience. With further refinement and upgrading of the product, JISHIYU is poised to deliver even greater value to clients across diverse industries."

CEO of JISHIYU Bai Tao

EqualOcean: Please introduce JISHIYU to us first.

JISHIYU Bai Tao: JISHIYU was founded in April 2021. Our company primarily focuses on self-developed AI technology, offering a unified intelligent customer service SaaS platform across multiple platforms and channels. Our main clientele comprises cross-border e-commerce sellers, aiming to meet the needs of these enterprises venturing abroad. Our services encompass customer acquisition, lead generation, customer operations, brand management, and more. The platform's functionalities cover tools for social media platforms like Facebook, Instagram, as well as chat plugins on official websites, including post-sales emails. We also cater to overseas clients with functions like automated replies through commonly used communication channels with sellers, customer management, order management, and VOC (Voice of the Customer) data management.

EqualOcean: The company profile of JISHIYU mentions that over 80% of the company members are still concentrated in the technological aspect. Were most of the initial founding members primarily from a technical background?

JISHIYU Bai Tao: Yes, quite so. The company's main partners are two individuals: myself and CTO Feng Yan. We were roommates during university, both having pursued our bachelor's and master's degrees in automation at Tsinghua University. After graduation, we both joined China Mobile, where I focused on the customer service department, specifically 10086. I led a team to establish the entire China Mobile CRM system and operational management processes. Later on, I began managing teams and oversaw the management of several thousand personnel in Beijing Mobile's 10086 department. Following that, I worked as VP at Yonyou Communications, responsible for marketing, sales, and CRM-related tasks. Later, a few friends and I founded Taiyuan Technology, specializing in outsourced customer service. Taiyuan Technology went public on NASDAQ in 2015. Therefore, for the past 20 years, I have been predominantly engaged in customer service and CRM-related work.

EqualOcean: It seems you recognized the potential of this sector early on, and your previous experience is closely aligned with it.

JISHIYU Bai Tao: Yes, because of my technical background and familiarity with CRM's operational experiences and processes, including platforms, I felt well-prepared. In terms of pure technical aspects, it's been many years since I've been actively involved; I've gravitated more towards management and front-end marketing. As for Feng Yan, he worked on the Feixin (Fetion) product at China Mobile. After leaving China Mobile, he became a co-founder of Wowo Group (55tuan), thus accumulating 20 years of experience in the internet industry focusing on product development and management. When we decided to embark on this venture, he primarily oversaw product development and AI, while I handled the company's management, sales, and marketing domains.

EqualOcean: While developing JISHIYU and its products, which aspect of improvement and enhancement do you emphasize the most?

JISHIYU Bai Tao: Our company was established in 2021. In 2020, I, along with a few friends, began exploring directions. At that time, the backdrop was the surge in online shopping demand in Europe and America due to the pandemic, propelling cross-border e-commerce in China like a "rocket." However, when we spoke with sellers, we discovered that they were grappling with a crucial issue. Communicating with global users, especially those from Europe and America, posed challenges due to time zone differences and language barriers. Even when using English, cultural disparities often impede smooth communication. It was evident that interactive engagement with users in online transactions was pivotal and core.

This led to what we term "service and marketing integration" becoming disjointed; inadequate service adversely impacted sales. Therefore, their demand in this area was substantial. This context prompted us to establish JISHIYU. Their pain points centered on time zone, language, and cultural disparities. Precise AI, available 24/7, effectively neutralizes time zone issues and can automatically detect languages, including cultural nuances and idiomatic expressions. Through continuous training and model refinement, we address these challenges.

When it comes to product development, our utmost focus is on integrating AI capabilities with cross-border e-commerce application scenarios. To illustrate, consider post-sales service. In domestic e-commerce, logistics inquiries are scarcely raised because of the well-developed domestic logistics networks. However, cross-border transactions entail numerous logistics-related issues. If your responses are delayed, users may request returns, or some users might be unsure how to use the products, prompting return requests. Drawing from my years of experience in customer service, I know that cross-border sellers attach great importance to emails regarding received items. As the quantity of items shipped is large, their cost per unit is low. Conversely, the cost of returns often exceeds the item's unit price, leading sellers to sometimes prefer refunding instead of accepting returns. This results in significant losses. In such cases, our AI, when combined with the scenario, automatically identifies users seeking returns. AI promptly sends an email to appease the user, "We have received your request. Please wait patiently. We will promptly assign someone to assist you." This single sentence extends the user's waiting time by four hours. Naturally, we don't actually make the user wait four hours; the AI identifies important emails and prioritizes them in the customer service representative's queue. Through this scenario alone, many customers' return and refund rates have decreased by over 30%.

EqualOcean: In this process, will AI identify different key terms and key phrases to prioritize sorting?

JISHIYU Bai Tao: We refer to this principle as intent. Just using key terms and keywords isn't very accurate. For instance, adding "not" before a word or phrase completely changes its meaning. Especially in the modern internet context, there are many slang terms and abbreviations, making it difficult for non-locals to understand. However, due to our AI's continuous annotation and learning efforts, we can handle these challenges.

EqualOcean: In the realm of intelligent customer service providers, what is the core competitive advantage of JishiYu compared to similar products?

JISHIYU Bai Tao: I believe our core competitive advantage lies in the fusion of AI and contextual understanding. Our comprehension of various scenarios is profound. Among domestic competitors, we were the pioneers. Since 2020, JishiYu has accumulated billions of language samples. We have a professional team of returnees dedicated to annotations, covering over 5 million samples. The accuracy of all user-accessible intents stands above 95%. Unlike other technologies, the accuracy of AI is easily recognizable. Hence, our AI capabilities combined with practical application and deep contextual understanding constitute our primary edge.

Another crucial strength lies in customer support and collaboration. Essentially, we are the first in the industry. Whether it's Facebook, email, or the current live chat, whenever we introduce a new scenario, users show significant interest and collaborate extensively. For instance, if we need language samples, they provide them; if we require testing, they assist. As the first in the market, we have access to these resources, giving us a substantial advantage in terms of timing and user cooperation.

The service provided by JISHIYU

EqualOcean: Speaking of AI, we must mention the wave of large models since the end of 2022. JishiYu's technology is closely related to natural language processing. How does JishiYu approach or view this AI large model trend?

JISHIYU Bai Tao: Firstly, we definitely embrace technology. Before the release of GPT-3.5 in November last year, we had been developing our own AI models. Our NLP is based on words, and we also involve images, GIFs, videos, and emojis, forming a multimodal model. These various modalities are all internally developed. Coupled with our profound industry understanding, when we officially launched our products in the second half of last year, we already possessed a highly precise GPT library. GPT's timing is perfect because our previous model accurately identified user intent and generated responses, but these were manually crafted, lacking diversity and standardization. With GPT now available, combined with our intent-based model, we can provide more nuanced scenarios, enhancing our service to sellers.

EqualOcean: GPT is a rather general model, but in customer service and specific contexts, it still needs tailored integration, right?

JISHIYU Bai Tao: Indeed, for example, if a customer asks, "I'm 170 cm tall and weigh 70 kg, what size should I wear?" GPT wouldn't be able to answer directly. However, in our context, we already have a knowledge graph about products from sellers. GPT is used to refine language samples, making them more human-friendly.

EqualOcean: JishiYu mainly serves cross-border e-commerce, targeting which international markets, and what is the approximate distribution?

JISHIYU Bai Tao: Our model is based on English, while other languages use our optimized translation models with underlying support from Google Translate, Bing, and GPT. Our support for European and American languages is more comprehensive. Presently, around 80% of our clients are focused on the European and American markets. The remaining portion includes Southeast Asia, South America, Japan, Korea, and the Middle East, but the proportions are lower.

EqualOcean: Has there been any change in this proportion from 2021 to now?

JISHIYU Bai Tao: The proportion on our platform hasn't changed significantly, as our target market remains clear, focusing primarily on Europe and the US. From the sellers' perspective, there have been some changes. For instance, some who were previously focused on Southeast Asia's COD are now shifting to COD in Eastern Europe. Similarly, South America has recently shown robust growth, and sellers specializing in the Middle East have developed rapidly over the past year or two.

EqualOcean: Why did JishiYu choose the United States as its first overseas office? What functions does this office primarily serve?

JISHIYU Bai Tao: Choosing the United States is partly related to our model. Since we work with English language models, it's natural that we would have better support for the English language system. Applying for overseas resources through a US company might be more convenient. In fact, we use interfaces provided by a US company for ChatGPT, including some US payment methods. Applying for overseas resources is one aspect, and the other is that our goal is to become a global company. The clients we have in China currently help us refine our products and improve our AI. Establishing a company in the United States is also beneficial in this regard.

EqualOcean: As you mentioned earlier, JishiYu aims to become a global company. In the company's introduction, we have seen that JishiYu plans to establish multiple wholly-owned subsidiaries or offices globally in the future. Which regions will be the main focus of this expansion?

JISHIYU Bai Tao: The company was founded just over two years ago, and in terms of products, our first priority is stable development. Future expansion will be influenced by two factors: continuous product iteration and improvement, and the overall environment. It's clear that the global environment in recent years is not as optimistic as before. However, the economy works in cycles, and there will definitely be upward fluctuations. When both our products and the overall environment are relatively ready, we will certainly expand. Our initial focus will be on Europe and the United States. First and foremost, the United States, followed by Europe and South America in terms of data. These regions, in terms of economic strength, purchasing power, and the prevalence of e-commerce, give us a fair amount of confidence. Additionally, our model's support for these languages will be stronger.

EqualOcean: Both domestic and international economic situations have undergone some changes in recent years. What changes have you noticed in the cross-border e-commerce sector?

JISHIYU Bai Tao: From 2015 to 2020, there was a trend of rapid growth, where people didn't pay much attention to operations and services. They used China's supply chain advantages to sell products, operating in a rather rudimentary way. However, this approach was not sustainable. Just like the development trend of domestic e-commerce, which started in 2015 when Alibaba began to regulate items on Taobao, domestic sellers began seeking operational tools. So, starting from 2020, we were deeply involved in this process and started setting up the framework in advance.

As expected, in 2021, there were various account suspensions for sellers, which compelled them to become more compliant. This involved ensuring compliance of products and also the compliance of operations. Sellers needed to focus on meticulous operations and brand building. The era of high-profit margins through simple and crude methods has passed. Tools are necessary to reduce costs and increase efficiency. We've clearly noticed the evolution. Initially, clients requested simple functionalities, like handling Facebook comments. Just responding to users was seen as a significant improvement. From last year, we've experienced a clear shift from rudimentary to refined practices. For example, initially, there weren't many clients asking about customer service quality checks. But now, larger sellers inquire about quality checks. Last year, we introduced a VOC feature, which stands for "voice of the customer." Through interactions with users, we assess whether they have any dissatisfaction with a merchant's products, services, or logistics. We have a module that automatically categorizes and provides reports based on this. Initially, this feature wasn't widely used, but this year, many sellers have started using it and providing various requirements to make it more sophisticated.

EqualOcean: Bai mentioned earlier that JishiYu now has the VOC feature, a new application. We've seen other service providers expanding into visual aspects. However, JishiYu's services are more closely related to natural language processing. Will JishiYu explore more projects related to computer vision in the future?

JISHIYU Bai Tao: In the short term, we may not consider these areas. Text-based AI research is our expertise, and we have invested many years into it. If clients have demands in other areas, given that there are already many digital and virtual teams on the market, we can collaborate with them. This is one aspect.

Additionally, as I mentioned earlier, my expertise lies in building a complete CRM system and process. Within CRM, there are elements of marketing, sales, and service. Integrating service and marketing is something I am very skilled at. Many companies focus on marketing and sales, and as we deepen our focus on service, it's natural that we can extend into marketing and sales scenarios. Establishing a platform that covers the entire process of customer relationship maintenance and management is the direction our product is headed in.

EqualOcean: When considering future business expansion, you're primarily thinking from a scenario perspective?

JISHIYU Bai Tao: Yes, and if I were to add another dimension, it would be by industry. Currently, we're in e-commerce, but many industry tools are transferable. In the future, we may consider serving industries like finance and gaming.