Conversation with FancyTech CEO Kong Jie: How does this AIGC company achieve annual revenue of more than 100 million in To B business?

Blue Whale News, February 19 (Reporter Wu Jingjing)Under the wave of AI models, how to make AI To B business sexy?

The commercialization performance of AIGC company FancyTech gives their answer.

In 2024, FancyTech’s revenue will exceed RMB 100 million, a year-on-year increase of 100% year-on-year in 2023. The number of customers exceeds 1000, of which KA-level customers account for more than half. Among China AIGC startups, this kind of commercial performance is rare.

FancyTech was established in 2020. It mainly uses its self-developed image and video models to provide AI-generated product pictures, videos and other marketing materials for KA brand customers in the FMCG and retail industries, helping brand customers achieve intelligent production and upgrade of marketing content.

The company also recently completed a US$10 million Series B+ financing, which was jointly led by Zhilin Capital and old shareholder Jinshajiang Venture Capital, and Guangyuan Capital served as the exclusive financial consultant. Zhu Xiaohu has also publicly praised this company many times and participated in investment three consecutive times.

In the four years since its establishment, FancyTech has gone through three transformations and embarked on a unique path on the commercialization of AI To B.

Unlike many AI model companies who first polish their technology and then find scenarios, FancyTech has taken a different approach and found its own commercialization path.

“Polishing the technology deeply before looking for business scenarios can lead to the wrong direction. rdquo; FancyTech CEO Air World spoke to Blue Whale News. In this wave of large-scale model entrepreneurship, the theory of technology first once prevailed, but FancyTech chose a path oriented by market demand.

In the commercialization process of To B, FancyTech mainly targets KA (Major Customer Logic) users, which is also the general perception of China To B companies in the past ten years: Compared with SMB (small and medium customers), KA customers have a stronger demand for digital management. The starting point for their use of tools is demand-oriented rather than technological upgrades, which also enhances the stickiness between tools and the companies using them, and brings more sustainable income.

However, in terms of specific paths, FancyTech has taken a completely different path. FancyTech has abandoned its project-based model that is tiring and heavy and has low profit margins, and has explored a more standardized path using the supply chain of AIgc content production. Build a platform:

·The demand side is controlled by the company, and all merchants are directly connected to FancyTech;

·There are two types of roles on the supply side of the platform. One is the workflow model that provides technical capabilities, and the other is the skilled talents who actually produce content.

·In the end, merchants only look at the effect, and the platform will be disassembled based on tasks and layered to producers and providers.

The picture shows FancyTech’s operation background

How to make this platform model truly flow?

For users, on the one hand, FancyTech brings new content effects to merchants through the platform, attracts more merchants to join, thereby attracting more model providers to settle in; for platform participants, FancyTech drives the business team to quickly test market demand and promote supply-side response and achieve scale replication, ensuring that technology providers can better understand market needs, thereby developing technologies and models that are more in line with merchants ‘needs.

“Similar to the physical supply chain, FancyTech focuses on the content creative supply chain, improves efficiency through AI technology, and serves global merchants. rdquo;Air Industry told Blue Whale News that this model can continuously bring new effects to the platform, attract more merchants, and encourage more model providers to join the platform.

FancyTech is trying to build an AI content ecosystem that connects technology providers and content creators, and drives it through business teams to meet the diverse content needs of KA merchants around the world. This not only responds to changing needs, but is also an inevitable choice to achieve scale.

This time,Blue Whale Technology talked to FancyTech CEO Sky World and chatted about their thoughts and practices on how AIGC products can find PMF and move towards scale.

Sky World graduated from the University of Nottingham in the UK and MBA from the Paris Business School in France. He was the head of content marketing for Alibaba’s Tmall Apparel and the founding head of Tmall Luxury Pavilion.

The following is a transcript of the dialogue, which has been edited and revised:

1. Redefine content production and create a content supply chain for AI content generation”

Blue Whale Technology: FancyTech was established in April 2020. Its original business was to help brands build big data intelligent application management systems. What stages has the company experienced in its development? How many transformations have you made?                  

Space:The company has mainly gone through three stages. In the first phase, from 2020 to 2022, AIGC has not yet emerged, and what it does isAggregate content direction。

In the early days of entrepreneurship in 2020, the opportunity we saw at that time was that some overseas platforms, such as Shopify, could provide components that allowed users to directly see the related content of products on Instagram or TikTok on the platform without having to jump. We believe that there is also such demand in China, such as the ability to directly display the content of Xiaohongshu or Douyin on Taobao stores or merchants ‘WeChat stores.

The second stage, 2022, coincides with the rise of AIGC. We found that many popular content accumulated in the past by merchants is very valuable. We can disassemble it, make data annotation, and then use this data to train the model to produce better results. After that, we trained the model ourselves.

It has now entered the third stage. Currently, the models on the market, especially open source models, are very rich. With this opportunity, we have transformed into aAI content platform, there are various models suitable for different types and styles on the platform to provide comprehensive services for merchants.

Blue Whale Technology: Why did you choose to build a platform?

Space:Building a platform is a gradual development process. During the commercialization process, the team found that as the needs of merchants became increasingly diverse, relying solely on its own strength was too limited, so other roles were introduced into the platform to meet the different needs of merchants.

Blue Whale Technology: Is the model self-developed?

Space:From 2023 to 2024, the team trained its own models and filed them in the National Large Model Algorithm Library. At the same time, it obtained patents in the United States and complied with ISO standards in Europe. This is what has been done in the second phase.

As we accumulated a high-quality customer base, the business team gradually matured. After entering the third stage, we began to develop platforms. On this platform, there are not only self-developed models, but also external models are introduced, including models purchased by us and those that can be accessed by third parties.

Users may use 5 models or 6 workflows on the platform, they put forward requirements, and we disassemble and layer them based on tasks on the platform, assign these tasks to the appropriate content producers and service providers.

Blue Whale Technology: How to understand the business model of our platform?

Space:It can be summarized as follows: First,KA merchants focusing on the To B field, customers include various domestic and foreign brandsSecond, we do not sell tools, but directly provide the final content services. Third, we are platform-based. There are model providers and demanders on the platform.

At present, all types of video models, whether they are major companies in the United States or China, such as Byte and Fast Hand, mostly start from To C (consumer-oriented) and provide models or tools to allow consumers to directly generate content. We focus on To B, which mainly serves the content needs of merchants. Some merchants need photos to display their products and sell goods, and some need to make promotional videos and post them on social media or offline channels. These are all professional content of To B.

Don’t focus on sales toolsIt’s the essential difference between us and companies that provide tools or SaaS services. Our platform can replace photographers, studios or studios, using AI to deliver the final content directly to merchants, without merchants having to edit it themselves.

The picture shows FancyTech customer background

I think this is the real AI service.If only one tool is provided and internal personnel of the merchant still need to edit it, then the quality of the content will largely depend on the person operating it, which is not AI enough.Therefore, we deliver the results directly.

How to achieve this process? The way is platformization.

Our AI platform has two types of roles, one is the service we invite.Service providers, who generate and compose content based on models on the platform and are responsible for actual content delivery。the other isTrainers and providers of various models,They will continuously tune and train models on the platform based on customer needs.

Platformization is very important. Like Disney, which is also a content exporter, can become a big company. There are popular IPs every year or even every quarter. In fact, most of these IPs are not original to them. The essence of what they buy is that they have achieved platformization and cooperated with many studios. Hot IP products are sometimes unpredictable and may be just small hits at first. They buy, customize movies, and then promote them through streaming media and theaters. Then, you test which products are more popular in Disney Mall, and then license these products to dealers around the world. So, what Disney ultimately earnsDifference between global sales and initial IP purchase price。In this way, they can also launch a steady stream of popular content.

The same is true for us, combining the company’s global business capabilities and business channels, quickly testing, and if merchants are satisfied with the results, they will quickly promote them. In this way, the platform can continuously bring better content to merchants, and good content effects will attract more merchants. At the same time, as more merchants join, more model providers will be willing to settle in.

Blue Whale Technology: In the early stages of the development of large models, some major manufacturers also claimed to build large model platforms, but some just set up a platform for everyone to access. In terms of platformization, how can FancyTech do more in-depth? What more diversified services and support can be provided to demand sides or technology providers?

Space:We have not just built a platform. FancyTech can be further commercialized because it builds“Semi-platform model, that is, we control the demand side,All merchants cooperate with us. The company has direct customer channels at home and abroad, and a business team can quickly collect merchant needs and test them.

this is a“Model of content supply chainIn the past, people often talked about the physical supply chain, and some used flexible supply chains to achieve rapid refund of orders. For example, SHEIN’s model is to gain insight into popular clothing styles at the beginning, and when seeing that a certain style may become popular, they would find a small factory in Guangzhou to quickly produce samples and launch products. Once they see overseas consumers placing large orders for a certain sample, they will immediately ask domestic factories to speed up the production of this product, greatly improving the efficiency of manufacturing explosive products.

We also adopt a similar model,By controlling the needs of merchants at home and abroad and realizing that a certain function may be popular, they will find models or workflows that can be made domestically, create demos, and then communicate with local customers through the global business team.Once the customer places an order, the domestic team will quickly make it. If more merchants are interested, quickly copy them in batches.

this model willThe key is to control demand。The demand must be large enough to support the corresponding size before the platform can truly operate. Therefore, we need to cooperate with enough merchants to quickly test market demand. Once the demand is clear, the supply side will supply quickly, so that the entire process can continue to roll.

Model developers are willing to cooperate with us because cooperation requires a strong fulcrum.If you just put the product on it but there is no traffic or users, then partners will not see the benefits and will naturally be unwilling to join.

On the FancyTech platform, when you put your product online, we can help you sell it. It can even be said that only when you see our successful sales will you be more willing to cooperate with us. Therefore, merchants are more willing to join hands with us. In this way, both sides can form a virtuous cycle. We did put a lot of energy into building our business team in the past.

Blue Whale Technology: SHEIN’s biggest advantage is the manufacturing advantage of China’s garment industry chain. By analogy, FancyTech is working on the content supply chain platform. What are the technical and product advantages of supply-side China?

Space:In the past, China had many factories with advantages in physical production. Similarly, there are now new opportunities in the field of content creation. China has the world’s largest talent market, and they face a variety of career options, such as traditional working from nine to five, or opening Didi and delivering takeout. But on our platform, they have the opportunity to do creative work.

In the past, the threshold for various software tools was high. Now, they can directly utilize AI models on the platform without having to learn photography, videography or complex editing skills. For example, in the past, individual creators did not understand Middle Eastern Arab culture and were unable to serve local users. Now, we collect data and conduct further training in different regions, and master the content styles, filter color preferences, and religious customs of the Arab region. These are the advantages of the system.

In addition, technically, the two major powers in the AI field are currently the United States and China. The United States may be more advanced in cutting-edge exploration of AI technology, while China performs well in AI applications. A single large model cannot cover all requirements, especially for enterprise scenarios, where every detail needs to be handled accurately. example,When a model changes clothes, the pattern, logo, matching jewelry, rings, etc. on the clothes need to be controlled separately. These cannot be achieved through a large model, but require workflow adjustments at the top.

In this regard, China has excellent technology, especially the many small personal studios that can achieve these fine adjustments. We introduced these resources to form a content creative supply chain.

2. Don’t let technology go astray: AI To B commercialization requires finding the right scenario, and technology first is a misunderstanding.

Blue Whale Technology: At the commercial level, domestic companies in the video model field include platform-based companies like Kuaishou, which have their own platforms and can assist creators. There are also some startups. Some are considering expanding To B (enterprise-oriented) application scenarios; others are targeting the ToC (consumer-oriented) market, such as focusing on the generation of film and television content. What is the essential difference between the video model business done by these companies and what we are doing?

Space:First of all, judging from the results, we are almost one of the best commercialized AI video startups in this round.

The company will achieve approximately 4 to 5 times growth in 2023 compared with 2022. The revenue in 2024 will be twice that of 2023, reaching US$15 million (over 100 million RMB), which is almost the highest among AI video startups in China.In 2025, we expect revenue to double again to US$30 million (over 200 million yuan), such achievements are also among the top in the world.

Why can such commercial results be achieved?

Because we provide more than just a SaaS tool, such as an online website or editor for merchants to use.Many of these tools may charge only a few dozen dollars a month or year overseas. What we provide is content, a service that replaces the original content production method.

The first concern for merchants is whether the content can achieve the quality and quality they expect, rather than efficiency or cost. Our core is also to help merchants improve the quality and efficiency of content production. So, technically, when we trained early models, we focused on physical restoration and control so that objects could be truly and vividly expressed, including details such as light, light, shadow, and material.

Now, we are expanding into a wider area. Overseas already has local customers in more than 10 countries, all of which are KA (Key Customer) merchants, from South Korea to Thailand, Indonesia, Singapore, United Arab Emirates, to Saudi Arabia, Europe, the United States, Brazil and other places. In Saudi Arabia, I also held an event with more than a dozen important local customers yesterday.

In addition to e-commerce, we have expanded more tracks overseas. such asretailUnlike many online retail stores in China, most overseas retail stores are still offline. Our content services replace offline billboards, such as video billboards on both sides of the road, restaurant menus, hotel promotional materials, and even areas involving advertising content such as entertainment, games, and financial banking services.

To sum up, our commercialization is divided into three steps: the first step isFocus on To B field, but the content itself is provided rather than the tools, and the criterion is the quality of the content; the second step isImprove production efficiency and achieve technology optimization, this is achieved vertically through the model we trained ourselves; the third step is, further platformization to introduce more role participation.

Blue Whale Technology: The core reason why the company’s commercialization has made very good progress does not depend on how strong your basic model capabilities are.

Space:It is impossible to summarize it simply in one sentence, because the commercialization process includes multiple stages.

First of all, we focus onVerification of product-market fit (PMF)。That is, how to enter the market? The first step is to start from scratch and strive to gain recognition from top customers. The second step is to achieve scale in a certain industry when the scale of the business reaches tens of millions. The third step is to think about how to expand to other industries, such as expanding from e-commerce retail to other fields, achieving a business scale of hundreds of millions.

The key to PMF is to find its own characteristics. Our specialty is to serve To B KA merchants, providing customized content delivery rather than just providing tools.When our clients work with advertisers, they need to submit creative briefs and then shoot and produce advertising content. This process involves many aspects such as creativity, refinement of content, consistency between color layout and brand style. We use AI to meet merchants ‘high standards for content and is more efficient than the original process, thus winning the trust of customers.

With the help of top customers, the next step is to expand the scale.For example, it has grown from millions to tens of millions. To this end, we used our own GPU to train the model and accumulate relevant data. The team has dozens of people dedicated to data annotation. Unlike some video models for To C, some of them focus more on breadth and generation capabilities, our model training focuses on authenticity. Only when model capabilities are closely related to business entry points can they support expanding business scale.

The company is also continuing to invest in technology research and development. In 2024, the team published 8 related papers, and the research in vertical fields has become more and more in-depth.I believe that we should not delve deeper into technology before looking for business scenarios, which may lead to going in the wrong direction.

Blue Whale Technology: First hone the technology in depth and then find the scene. This is the path many video models are taking.

Space: This may cause a wrong route. We chose to cut into the scene first to find requirements and verify them, and then deepen our own technology in the second step.The overall situation is also moving relatively fast. Starting from the second half of 2024, we have already begun to promote platform construction.

Blue Whale Technology: In commercial choices, have you made detours?

Space:I tried to create the SMB market for small and medium-sized businesses, but I couldn’t start a lot. I still returned to KA merchants, which may be related to the team’s genes.

Blue Whale Technology: In terms of specific content generation, how to avoid the problems caused by machine illusions?

Space:There was a time when it was widely believed that one large model could solve all problems. But now it seems that this is impossible. Therefore, a platform-based solution is needed. The large model can be adjusted accordingly according to each type of specific needs, which may be achieved through fine-tuning, using the LoRA method, or through refined control of workflow.

Initially, we developed our own model and did a lot of tuning work on it, including self-training and data accumulation. After discovering that this path was feasible, we replicated and expanded the method from one model to ten or a hundred. Now, we have implemented a platform, inviting different users to reuse our method, and promoting and amplifying it through our business team.

3. How to scale up the sea? Localized operations are the key

Blue Whale Technology: You have been traveling overseas in recent months, and your current energy is mainly focused on going overseas?

Space:The company’s strategy in 2024 is to start from the two fast-growing regions: the Middle East (especially the United Arab Emirates and Saudi Arabia) and Southeast Asia (mainly Thailand and Indonesia), hoping to see results quickly.

Next, it will further expand to other regions, such as Japan, South Korea, Europe and the United States. Although the development speed of these places may be slower, when we accumulate certain experience and foundation, we can go further.

We have accumulated certain popularity and influence in overseas markets. In 2024, we have achieved remarkable results in Europe and won the first place in the highest award at Viva Tech, the largest local technology exhibition. At the same time, we have also won the title of Forbes Asia ‘s 100 Most Watchworthy Companies.

Blue Whale Technology: Overseas, such as the Middle East and Europe and the United States, there are actually significant differences in aesthetics and culture. How to meet different user needs?

Space:Every region requires heavy initial work in the early stages.Taking the Middle East as an example, we need to understand and learn from local culture through local teams.Local data also needs to be collected. If the requirements are higher, we will even set up cloud computing and training nodes locally, conduct further training and collection based on local data to meet specific local needs. Even as long as we have core customers locally, we will establish a system locally to meet local data security requirements, data characteristics, and local regulations.

Once these tasks are established, they constitute our barriers. As mentioned earlier, delivery content itself is a barrier that distinguishes us from tool companies in the United States. Because their tools still need to be used by people, and we can provide services that are more suitable for local needs. Also,It is also impossible for American companies to optimize specifically for different styles, such as setting up training nodes specifically for the Arab region. They do not have such resources and energy. This is precisely our advantage.

When we accumulate enough experience in this area, we will be able to better serve customers and attract more customers. As technologies and models increase, more merchants will be attracted to join. The scale can also be further expanded in terms of commercialization.

Blue Whale Technology: Have you summarized a set of reusable market development methods that can be effectively applied in different countries or regions, whether from a commercialization perspective, technology implementation, or other aspects?

Space:How to enter the market and expand. This process also needs to be carried out in stages, which can be roughly divided into three stages.

Getting into the market at first is very difficult.Fortunately, we seized the opportunity in 2024. At that time, major cloud vendors were very eager to promote AI functions or solutions.However, cloud vendors mainly provide basic services, and merchants often do not know how to apply these services. At that time, weWhen promoting overseas, cloud vendors are very eager to incorporate our products into their solutions. Because our products are very intuitive, customers can understand how to use them at a glance.

So, in the first phase, we were able to quickly launch in 10 countries, mainly thanks to the help of cloud vendors. Moreover, this process does not require complex layering. As long as it is successfully promoted, local cloud services will naturally be used.

After entering the market, the second stage isLocal recruitment of business staff。The content production side mentioned earlier is placed in China, taking advantage of China’s technological dividends and personnel advantages. At the same time, we also use demographic lipsticks in countries such as India and Egypt, for example. Talents from these countries also have deep penetration in overseas markets. For example, Indians are common in many companies. As a result, we are able to quickly recruit local business people and help them quickly enter the market, complete conversions and verification.

Now we are entering the third stage. There is already a local business team and continues to carry out business. Next, the scale of the business will be expanded through local channels. Overseas, it is usually the SDR (Sales Development Representative) who helps us obtain clues and introduce customers, and the final quotation process is completed by our own business team. This can quickly expand the size of the business team. This is also the standard To B market approach overseas.

Blue Whale Technology: To B industry, the overall ecological chain and sales culture abroad has a very standardized process and architecture. This is very different from the domestic highly customized or specific KA logic. How did you first decide to choose To B sailing direction?

Space:We started from the domestic market and have achieved a leading position in the industry in China. At that time, some of the international brands we worked with also had business overseas, which became our earliest opportunity to showcase our strength overseas.

After gaining successful experience, the company decided to fully launch overseas expansion plans. This is a very natural development process.

Blue Whale Technology: The company’s customer categories include clothing, beauty, and luxury goods. Is there a big difference in the proportion of different categories?

Space: If we only look at it from a retail perspective, our proportion in each core category is relatively balanced, such as 3Cs, digital products, FMCG, food, drinks, and clothing, and sporting goods. In overseas markets, in addition to retail, there are many other industry needs. For example, the offline service industry mentioned just now includes hotels, restaurants, finance, banking and other industries, as well as the real estate industry.

In some industries, products need to be pre-sold, and real shots are unrealistic, so content can be created through AI.In addition, there are entertainment industries, such as games, and local cultural industries, such as tourist attractions. At present, we are involved in basically all industries that need to produce advertisements in overseas markets.

However, our user profiles are actually very clear. They mainly serve leading large companies or KA merchants, rather than SMBs (small and medium-sized merchants). They are basically medium to large-scale merchants.

Blue Whale Technology: Give an example to illustrate how the company’s AI platform helps users achieve an improvement in the overall input-output ratio?

Space:Let me give some examples from different regions. In Thailand, we partnered with Villa Market, a local fresh food market. They do both online and offline business. In the past, when they sold bananas and mangoes, they could only use white background pictures. Now, we can automatically generate more attractive pictures and videos for them, such as the scene of a banana hanging on a tree, the growing process of a mango, and even the cooking process of cod from raw to ripe, just like a video that teaches you how to cook. Thereby saving their shooting costs.This is not just a cost issue, but also a process of creating content from scratch.

Another example is in United Arab Emirates. Real estate in the United Arab Emirates is very popular. There is a very famous local group called Landmark. They do both real estate and home furnishing business. In the past, home decoration required professional decoration companies and designers. Now through AI technology, this process can be made simpler. AI can help combine different household items and adapt them to different apartments, even if these apartments have not yet been built.

This content not only increases the user experience, but also helpsimprove sales。For example, some overseas brands require a large amount of material to publish on social media or Google Ads. In the past, due to limited manpower, it was impossible to produce so much content in one day. But now, AI can be used to generate large amounts of material, and if AI can connect to data, it can also generate different content based on different groups of people. This is a double improvement in efficiency and effectiveness. The content seen by different regions and different users will be different, which can only be achieved by AI.

Blue Whale Technology: Going abroad means that competitors come from all over the world. Compared with overseas manufacturers, what are the company’s core advantages?

Space:We should make full use of China’s advantages. Just like the United States, although it may be more advanced in some single technologies. But they usually make it lighter, focus on a certain function or model, and build a tool-based website based on this for everyone to use. Doing so will inevitably lead to their inability to solve overall problems, especially in the enterprise-oriented (To B) field, which will be limited because the To B field is relatively complex and requires more mature.

They are also not effectively delivered, tools may need to be run multiple times,For example, a tool may need to be run 10 times to get a usable result once, and each run may need to wait 20 minutes. Such efficiency simply cannot meet To B’s needs.

In terms of business model, we focus on delivering results, and in the early stages, we do it in depth. We have made full use of all parts of China that can optimize technology, and have also mobilized the power of a large number of content creators, which is not available in the United States or elsewhere.

In addition, we attach great importance to everyThe localization work in various regions is committed to creating solutions that suit local styles and accumulating local data for training.This is a gradual and cumulative process that allows us to do more and more in-depth and professional in enterprise-oriented services (To B). This in-depth and professionalism constitute our barriers and values.

Other startups may panic, such as when OpenAI launches new features, a group of startups may be replaced because they haven’t gone deep enough in the first place. I’m not worried about that right now.

After entering the third stage of platformization, our position has become difficult to shake. Because whenever a new technology emerges, we can quickly introduce it into the platform through cooperation with our partners, thereby further improving efficiency. In turn, it also gives the company’s gross profit margin room for further improvement. This strategy allows us to continue to enjoy the dividends brought by the AI era.

Blue Whale Technology: The current number of customers of the company? What is the proportion of domestic and overseas business?

Space:At present, the number of our customers should have exceededof 1000Among them, there are hundreds of customers at KA level.

In 2024, our business focus will still be domestic, and the ratio of domestic and overseas business will be about 7:3. We hope to achieve the proportion of domestic and overseas business in 2025.5:5。

Blue Whale Technology: What are the company’s next development plans?

Space:In terms of product technology, our models and platforms need to be further enriched and improved, and we are actively inviting more partners to join and continuously expanding our scale.

Secondly, at the commercial level, our business team plans to further expand. Especially in overseas markets.

Moreover, we need further financing. At present, Sino-US relations have indeed had a certain impact on our financing. Compared with foreign companies, our financing amount is relatively small with the same number of customers and revenue levels. In China, the main ones that currently have more financing are former large model companies. For commercial companies, some overseas cases show that even if they only have 200 or hundreds of customers, they can raise tens of millions or even nearly 100 million US dollars in the global To B market (such as law, sales automation and other fields). Their customer unit price and sales scale may be comparable to ours, but the financing amount is much higher than ours. This is the direction we need to work hard to break through in the next step.