The overlooked driving school business: They have earned hundreds of millions with AI robots | The "Little Giants" of the era.
Introduction: With the advent of the era of hard technology, a large number of hidden champions, single-item champions, and specialized, refined, distinctive, and innovative small giant enterprises are beginning to emerge. These enterprises are the core pillars of domestic substitution and the reservoir for Chinese companies to become global companies.
Based on this, Yingke has launched a new column: The "Small Giants" of the Era. Through in-depth communication with founders and investors, more small giant enterprises can be seen and recorded.
The following is the third article in this column.
Author | Wu Ruoyu
Editor | Peng Xiaoqiu
The huge imagination brought by AI has led more people to focus on large industries and big businesses.
Recently, Yingke discovered an interesting small giant enterprise that uses AI to replace driving school instructors, and its income in 2024 has exceeded 100 million yuan. Moreover, at the beginning of this year, it successfully received an investment from the well-known institution Dachen.
The driving school business is very traditional and is more likely to be overlooked. After all, a wave of Internet driving schools in China has already collapsed, such as the once popular Internet driving schools "Zhu Jianqiang" and "YY Xueche". However, no matter how strong the technology is, it ultimately needs to be implemented and commercialized. Therefore, there should be no so-called contempt chain in the application of AI scenarios.
This company, called Yixian Intelligence, was established as early as 2015 and has currently covered 29 provinces (municipalities and autonomous regions) across the country. There are more than 6,000 intelligent robot vehicles in operation, with a market share of 93.6%. But for the founder, Ma Hong, "The current intelligent driving training industry is still in the early stage, and the story of Yixian Intelligence has just begun."
In fact, Ma Hong operated 9 driving schools in 2014 and is an insider. Years of experience in the industry have allowed him to see the pain points of the domestic driving training industry and has always hoped to use autonomous driving to solve the problems of non-standard and non-standard driving training scenarios. Therefore, Ma Hong collaborated with Professor Duan Guijiang from Beihang University to develop a prototype. After testing, it was found that the technical route could be feasible.
However, the advancement of intelligent driving training is not smooth.
The demands of listed companies have accelerated the commercialization of products and gradually increased the market acceptance. In 2021, the revenue of Yixian Intelligence has exceeded 100 million yuan.
After the novelty wore off, problems such as incomplete application scenarios of the product were exposed. Coupled with the impact of the epidemic, in terms of revenue, Yixian Intelligence experienced a trough period from 2022 to 2023, with the worst time dropping to 20 million yuan.
"This is an inevitable stage for technology-driven companies. When we promote new technology products from the 'early adopters' to the 'early majority' stage, we need to bear the corresponding risks and costs," Ma Hong told Yingke.
Ma Hong and the team spent one year to continue improving the product, compensating for the scene defect and technical insufficiency of only the second subject, and fully understanding the business scene. The persistence in the strategic direction has enabled Yixian Intelligence to resume growth. Especially the 9.0 product released in November 2024, the annual revenue has once again exceeded 100 million yuan.
Recently, Ma Hong shared with Yingke how he discovered this small but beautiful segmented track, how to expand the business step by step, and his thoughts on the transformation and development of the driving training industry. The following is part of the conversation (edited):
"Maladaptation" is the "Anxiety Disorder" of Industrial Transformation
Yingke: There are not a few cases of the combination of AI technology and scene application. Why did Yixian Intelligence choose the driving training scene?
Ma Hong: There are three points in total:
First, the pain points in the driving training market are more prominent. The market supply and demand are unbalanced, the industry management is insufficient, and the quality of instructors and the needs of trainees do not match, which requires targeted solutions;
Second, we previously conducted tests on the prototype of the autonomous driving vehicle in collaboration with Beihang University and found that this technical path can be feasible, and the commercialization of application-oriented AI in the driving training field can achieve a closed loop;
Third, the Asian driving training market has broad prospects. Currently, the number of driver's licenses held by the 1.4 billion population in China is about 35.7%, compared with the 70% in developed countries, there is a large room for improvement.
In addition, from a practical perspective, the main cause of traffic accident casualties is the non-standard behavior of drivers. We hope to reduce the probability of traffic accidents through intelligent driving training.
Yingke: I learned that you have previously founded 9 driving schools. Why didn't you continue?
Ma Hong: At that time, I wanted to create a chain of driving training brands. However, when it developed to the 9th one, I found that the standardization and suitability of the instructor team were insufficient, which could not be solved by exam-oriented education. Therefore, I considered using AI to replace humans, strengthen scene teaching, blur the boundaries and sequence of the exam, and help candidates understand a more real driving test.
Yingke: What stage is Yixian Intelligence currently at? In your opinion, is the acceptance of intelligent driving training in the industry high?
Ma Hong: Yixian Intelligence has now entered the early mass acceptance stage, and the orders from the leading driving schools in each province have increased significantly. It should be said that after 9 years of technical accumulation and scene demand exploration, the acceptance of our products is relatively high.
And I believe that in the AI wave, not embracing AI will not be able to survive. This sign is clear.
Yingke: Will the integration of traditional driving schools with AI technology lead to "maladaptation" in terms of business, personnel, and prices?
Ma Hong: Here, we should talk about the "maladaptation" of the entire industry. For the industry to transform and upgrade, it needs to get rid of the past profit model and business habits. This process is painful.
At the beginning, we were also confused and anxious. However, by accumulating experience through time and projects, we gradually ran through the entire solution process, and then fed back the cases to the products. I feel that we are a bit like Yu Chengdong making cars, participating in the entire process from the front end to the back end, providing more comprehensive and practical solutions for traditional driving schools. Now, when we deliver projects, they can be fully operational within basically one month, solving the problem of "maladaptation".
Secondly, there is the price issue. The economic downturn has led the entire industry to compete on low prices, and AI has come at the right time. For example, buying 1 robot instructor can replace 10 manual instructors in the past, saving the cost of 9 instructor workers per year. Considering the comprehensive equipment investment, the driving school can recover the cost in one year. Then, the decrease in cost is reflected in the trainee side, and the decrease in the learning cost is also more in line with the concept of consumers pursuing cost-effectiveness now.
From the Demographic Dividend to the AI Trend
Yingke: What are the main products of Yixian Intelligence currently?
Ma Hong: Currently, we mainly focus on the intelligent driving training system robot instructor RoboCoach®, which is divided into high, medium, and low configuration versions. And the pre-installed version of the intelligent connected driving training vehicle.
Yingke: What difficulties have been encountered in the research and development stage of intelligent driving training products?
Ma Hong: We have encountered two difficulties: First, safety guarantee. Autonomous driving perception, evaluation, judgment, decision-making, execution, and feedback can all achieve a self-closed loop. However, in the actual scene of intelligent driving training, after adding the "human" as an uncertain factor, the algorithm needs to be repeatedly debugged to ensure the safety of trainees.
Second, the vertical teaching model. We need to make the teaching model able to cover the entire scene of driving training. Previously, there was a lack of data, and we have stepped on many pits in the annotation of scene data knowledge. After nine years of exploration, our GoDrive large model can fully adapt to the business scene.
Yingke: How do you position Yixian Intelligence?
Ma Hong: We are an overall solution provider for AI mobile machine services in specific scenarios. I believe that the product ultimately needs to be implemented. Many startups focus too much on technology and use AI technology to promote business, making it difficult to achieve a perfect integration of products and scenarios.
Yingke: Who are the competitors of Yixian Intelligence, and what are their volumes and scales?
Ma Hong: In this segmented track, we currently have no competitors. Currently, there are some imitations in the market that will only harm customers. There will be competitors in the future, but they may not be in the driving training industry.
Yingke: The "Overall Solution for the Digital and Intelligent Transformation of the Driving Training Industry" proposed by Yixian Intelligence, how is it specifically implemented?
Ma Hong: Our solution is an implementable strategic execution plan. Yixian will, like Yu Chengdong making cars, deeply intervene in the entire process of customer operations, involving every link and contact point from leads, transactions, training, to exams and obtaining certificates, to help customers complete business operations.
Yingke: As the core product of Yixian Intelligence, what is the penetration rate of the RoboCoach® intelligent robot instructor? How to improve the penetration rate in the future?
Ma Hong: The current penetration rate is still very low, and it is in the early stage of industrial transformation from a niche market to a mass market.
We will take AI as the leader, drive the new energy driving training vehicles for the industry, build an online live broadcast matrix to reduce customer acquisition costs, and further build a new intelligent, green, and integrated business format for customers. Instead of copying the traditional Internet driving training customer acquisition method, we have jointly built a marketing matrix "Deshou Jiayuan" with customers with stable service quality control to improve the customer acquisition efficiency through Douyin's local life. Currently, the daily transaction amount has exceeded 10 million yuan.
Shuffling, Focusing, Waiting for the Emergence of New Giants
Yingke: What is the volume of Yixian Intelligence's downstream customers? How is the purchasing power and development space of its core customers?
Ma Hong: Our customer group is mainly medium-sized driving schools with about 10,000 people.
Under the original license system, the market competition is not fierce. Although there are more than 20 million people learning to drive every year, the market share of a single driving training institution does not exceed 2%. Nowadays, the driving training industry implements the policy of "Streamlining Administration, Delegating Powers, Improving Regulation and Services", that is, the filing system. Driven by the policy, the entire industry will be reshuffled to form new regional leaders. What we need to do is to identify and support customers.
I believe that within the next five years, 20,000 driving schools in China will undergo industrial restructuring through methods such as withdrawal and mergers. The core customers may remain within 2,000 to 3,000, and there will be leading brand driving schools with an annual training capacity of over one million people.
Yingke: What is the production mode of Yixian Intelligence? Is there an expansion plan?
Ma Hong: Currently, we mostly purchase shelf products for hardware, and the core domain control is designed and custom-produced by us, and then integrated into the production line for packaging and shipping according to orders. Therefore, the product form is highly flexible, and we will not make an expansion plan when the capacity is increased by 1 - 2 times. Currently, we have signed a pre-installed driving training vehicle with Dongfeng Automobile Factory and applied for the announcement of the intelligent connected driving training vehicle. It is expected to achieve vehicle delivery in the second half of 2025.
Yingke: Yixian Intelligence has a high domestic market share. What is the next plan?
Ma Hong: The next step is mainly to increase the penetration rate. There are more than 900,000 existing driving training vehicles, and at least 200,000 should be installed to cover 10 million people. This is the goal for the next 5 years.
Yingke: In recent years, AI driving training has also attracted much attention from investors. In the process of the industry moving towards explosive growth, what do you think are the urgent problems to be solved in the intelligent application and integration of the driving training industry?
Ma Hong: The urgent problems to be solved are institutional obstacles and financing support. On the one hand, we are promoting the release of industry standards and the implementation of policies. On the other hand, we are actively contacting financial institutions.
Yingke: Is there an overseas expansion plan?
Ma Hong: We are also paying attention to the foreign market, but we have not actively gone overseas. The main reason is that the domestic penetration rate is still insufficient, and I don't want to be distracted. If we consider the overseas market, it will also be centered on Asia.
Yingke: This year, autonomous driving products have continued to heat up. Suppose one day autonomous driving becomes very mature. Do you think people still need to get a driver's license?
Ma Hong: In fact, we have also been paying attention to this problem. The biggest concern about autonomous driving is safety.
First, I don't think autonomous driving is absolutely safe. Autonomous driving operates based on scene data, but the scene data is constantly updated and changing. Machines cannot learn 100%, and even a 0.01% gap may endanger the safety of passengers.
After systematic driving training and completely relying on autonomous driving without knowing how to drive, their safety upper limits are different. As long as autonomous driving has not surpassed humans, it is necessary to learn to drive.
Secondly, once fully autonomous driving appears in the entire society, problems such as privacy leakage and machine out of control will arise. Finally,