Back to Day 0 of starting a business.
Text: "Chinese Entrepreneur" reporter Yan Junwen
Editor|Li Wei
Header image source: Respondents
With the rapid change of technology, technology unicorns are ushering in a major reshuffle, and transformation or non-transformation will usher in a battle of life and death.
In February this year, a unicorn company called Emotibot in Shanghai reported that it was "suspended for 6 months". The company later issued a statement denying the shutdown but acknowledged layoffs and cutting lines of business.
Emotibot's specific reduction actions include: the number of products is directly cut by two-thirds from 12, and the remaining 4 products can be profitable; The number of employees in the company has shrunk from more than 300 at its peak to more than 100; The office area has dropped from more than 2,000 square meters to hundreds of square meters.
"We have to go back to the 0th day of entrepreneurship, strengthen the entrepreneurial spirit, and get rid of all the baggage." Jian Renxian, the founder of Emotibot, said in an exclusive interview with "Chinese Entrepreneur" recently.
Founded in 2015, Emotibot has 7 rounds of financing, with a total amount of more than 1 billion yuan and a valuation of more than 1 billion US dollars (about 7.2 billion yuan), becoming a unicorn. Its core strength is NLP (Natural Language Processing), and Jian Renxian was the vice president of Microsoft (Asia) Internet Engineering Academy before starting his business, focusing on search, Bing advertising and other businesses.
In the second half of 2023, it launched the large model product matrix "1+4", a low-level large model and four platforms for dialogue, training, knowledge and generation, and more than a dozen products, trying to "solve the last mile landing of the large model".
But soon, the transformation turned out to be a huge miscalculation - many new products did not make their own blood, customers did not recognize, it only brought losses and dragged down the company's morale. Jian Renxian's attitude towards large models has also changed from "all in" to "half in", focusing on marking products rather than innovative products, and putting profit first.
Transforming is no easier than starting a business all over again. At present, although Emotiput, which is in self-help, has not yet completely got rid of the predicament, Jian Renxian summed up three experiences in the transformation process:
Drafting: Jiang Chenyu
Go to the vast third- and fourth-tier cities
This year, Jian Renxian moved the company from Fenglin Science Park in Shanghai's Xuhui District to a nearby startup incubator and turned it back into a startup. Lilith is downstairs from the office, and the building next to it is Youzu Network, and miHoYo is also in a building not far away.
A large part of the money-making model of these local game companies in Shanghai depends on going overseas - miHoYo has been dominating the revenue of overseas games all year round, and Lilith is also the top three players in the game's overseas market. Unlike these game companies, Emotibot has focused its greatest efforts on China, especially in the vast number of third- and fourth-tier sinking cities, because Jian Renxian found that the AI needs there are far from being met.
Source: Interviewee
Jian Renxian found that third- and fourth-tier cities accept new technologies faster, because large enterprises in first-tier cities have to report layer by layer, find 10 suppliers to compare, and the repayment cycle can easily take two or three years, while customers in third- and fourth-tier cities do not need to compare, use if they are valuable, and stop if they are not.
"The best way to validate your product is to sink." Jian Renxian said.
For example, a third-tier city labor dispatch company approached Jian Renxian and asked if he could use AI to do live recruitment, Jian Renxian said that it could be done, and it could be directly made with Emotibot's X-coach products, without customization.
X-coach is one of the few products left behind after Emotibot's business was downsized. At present, Emotibot's products are based on the biggest goal of hematopoiesis and profitability, and are mainly based on AI training and AI assistants, such as AI recruitment interviews, which are the most focused scenarios of Emotibot, and they have independently developed an HR recruitment model, which can achieve a matching rate of 90% between resumes and positions.
However, although customers in third- and fourth-tier cities are enthusiastic, they have low budgets. For example, a large company in Beijing, Shanghai, Guangzhou, and Shenzhen can give a budget of 50 million yuan for an intelligent customer service, but the company in a third- or fourth-tier city may only have a budget of 300,000 yuan, and it requires that data cannot flow out and do not connect to the API of the closed-source large model.
"When the enterprise encounters this kind of problem, our value comes out, and this is a core point we have found, and the customer wants a cheap and easy-to-use solution." Jian Renxian said.
As for whether the base of these products is an open-source or closed-source large model, Jian Renxian believes that it does not matter, and Emotibot has developed a small model that is different from the large model for enterprises to call.
From this point of view, Jian Renxian is a loyal believer of Zhu Xiaohu, the managing partner of GSR Ventures, who is completely doing to B, investing in projects that "can be commercialized immediately and basically do not need to burn money".
But Jian Renxian seems to disagree with Zhu Xiaohu's pragmatism, he said: "I have the ability to develop the basic model, I will do the basic model, whether I can raise capital is another matter, no one to do a thing he thinks is wrong in order to get financing." ”
"If investors take realistic returns as the highest principle and do everything according to the logic of investors, then there will be no progress in this society, and life will become meaningless." Jian Renxian sighed.
But idealism and technocracy are blind anachronisms for Jian Renxian in the present, and only by surviving can the future be turned around.
Jian Renxian said that when you are faced with a huge difficulty to break through, you must think of a solution with a completely different thinking than before, which can allow you to jump out of the original predicament and find a correct path, which usually requires overturning all previous ideas, all logic, and all processes, which is completely contrary to the comfort that the current team can accept.
"Of course, transformation is also against the comfort zone of a founder, and for a CEO, the real evolution has just begun." Jian Renxian said frankly.
CEOs should use anxiety as a tonic
Looking back at the entrepreneurial process from 2015 to the present, Jian Renxian recalled that he had always been worried that the speed was not fast enough, but now looking back, he feels that the speed should not be too fast, "don't be too anxious, don't be carried away by external information, don't be taken away by investors".
"If I had to make another choice now, I wouldn't dare to expand blindly, I would do as much as I could, and make a profit on what you do before moving forward, which is very important." Jian Renxian said.
The last time Jian Renxian received financing from investors was at the end of 2022, and the amount raised was not disclosed, with a joint investment from Gimpo Investment, Treasury Capital, Jiangsu Cultural Investment, and Junci Capital.
At this time, OpenAI in Silicon Valley released ChatGPT, a large model based on GPT3.5, and a new technological revolution officially kicked off. In the months that followed, the large model shook the tech world like a volcano. After getting a new round of financing, Jian Renxian also began to turn.
In July 2023, Emotibot launched the "Family Bucket" product of large model + product in one go, and when large companies and startups were still struggling with whether to make a large model, whether to B or to C first, Jian Renxian decided to do both.
This exposes a certain "ByteDance" mentality of unicorn companies, betting on dozens of projects in search of explosive models, but in exchange for endless losses.
This is Jian Renxian's "weeping blood" summary of the company's transformation in 2023 as CEO:
Drafting: Jiang Chenyu
Taking sales as an example, Emotibot has suffered losses. Previously, Emotibutial had set up a sales team of more than 40 people, but it was not very useful because the salespeople did not understand large-scale products. Jian Renxian said that the sales of traditional software, hardware and services in China basically do not meet the requirements of large-scale model sales. In Europe and the United States and other countries, the senior software sales staff of to B company have graduated from famous universities majoring in computer science.
So Jian Renxian went to the client himself, and he gave himself a full schedule - in mid-April, he had just returned from a business trip in Guangzhou, and before that, he met with clients and investors in Beijing, and in the following May, he traveled almost every week.
Faced with the downsizing of the business, anxiety inevitably struck Kan In-hyun. Jian Renxian said that he has been anxious since the first day of his business, and now he is anxious every day, how to solve it? "There is no way to solve it, let time solve it, let the effort solve it."
When asked why he didn't bring in professional managers to break down anxiety and share the pressure, Kan replied that finding someone to help you do the work will not solve your anxiety.
He then said, use anxiety as a tonic, after solving an anxiety, you are equivalent to taking a tonic, if you have indigestion, you have to find a way.
Is the current transformation of Emotibot the right one? Jian Renxian seemed a little confused: "I don't know either. He said, make a decision, do it, move forward, stick to the direction, solve the difficulties that can be solved, and change the direction if you can't solve them, there is nothing to worry about.
Lost and Rising Stars
In 2021, Jian Renxian was also a guest at various AI conferences, and he predicted that knowledge graph will be a key technology for cognitive intelligence in the next three years, and natural language technology is the bottom layer that supports knowledge graph. At that time, he was still the darling of capital chasing, and he could get a financing almost every year.
However, the knowledge graph has long become a thing of the past, and the large model is the most popular new outlet for investors. Obviously, Emotibot is no longer a popular entrepreneurial star among investors.
In Shanghai's Xuhui District, within a few kilometers of Emotibot's surroundings, MiniMax founded by Yan Junjie and Leap Star founded by Jiang Daxin are the hottest AI unicorns at the moment. MiniMax has received support from investment institutions such as miHoYo and Tencent, and ByteDance had planned to invest in Leap Stars, although it failed to achieve, but it can enter Zhang Yiming's vision, and the founders Jiang Daxin and Leap Stars have extraordinary natural abilities.
However, Yan Junjie and Jiang Daxin are trying their best to avoid the "tragedy" that happened to Jian Renxian.
Yan Junjie said in an interview with the media that for startups, if you don't have good enough product capabilities to undertake, even if you have some technical progress, these things are not yours in the end. A startup that develops independently must consider the product.
A friend once told Yan Junjie that many successful companies, such as miHoYo, Meituan, Byte, and Ideal, have one thing in common - none of them succeeded by the first product, but by the second or later products.
Headquartered in Caohejing Development Zone, Xuhui District, Shanghai, just a few hundred meters from Jian Renxian's startup, the veteran AI giant SenseTime is also thrilling in the face of generative AI.
According to the 2023 annual report, the total revenue of SenseTime's generative AI business in 2023 will be 1.2 billion yuan, accounting for 35% of the revenue, but on the whole, the annual performance of 2023 will achieve annual revenue of 3.406 billion yuan, a year-on-year decrease of 10.6%; The annual loss was 6.494 billion yuan, an increase of 6.6% year-on-year.
Xu Li, chairman and CEO of SenseTime, said that the decline in full-year revenue was mainly due to industrial transformation and the active contraction of smart city-related businesses in the traditional AI business.
This is also a strange circle that is difficult for technology entrepreneurs to break through, benefiting from technical barriers, but also trapped in technology.
An investor once told "Chinese Entrepreneurs" that when entrepreneurs have highly competitive technology, they should make good use of capital to further expand the leading position of technology, and capital is also one of the means of competition.
But this investor still hasn't said half a sentence, capital will only flow to better entrepreneurs, only see the new people laughing, not the old people crying, this is the attribute of capital, but also the rules of business.