Article at a Glance

Here's how Classting, founded in 2012, managed to survive and capitalize on this market before turning profitable with paid AI software-as-a-service sales in late 2023.

1. 10 years ago, they set a clear goal of eliminating the pain points of teachers owing to the limitations of one-to-many education and realizing "personalized education", so they opened their eyes to the potential of AI as a tool early on.

2. Acknowledging the reality of the slow-growth public education business, they have identified investors who are willing to wait, and they have developed a short-term revenue model that does not compromise the essence of education while raising funds for long-term investment.

3. Built "credibility" in public education by persistently convincing change-conservative teachers and school boards to digitally transform their classrooms, and by offering free services that prioritized teacher convenience.

4. Accumulated all "big data" related to a person's education, from learning data such as student problem solving and assignments to social data such as discussions and comments, and partnered with content companies to integrate the data for AI training.

---

When artificial intelligence (AI) enters the classroom, the landscape of the classroom will change: students who don't understand a lesson will be able to follow along with tweezer exercises and video lectures instead of letting it slide, and they'll be able to retrace their steps and re-learn the basics. To realize this blueprint for personalized learning, leading edutech companies and textbook developers are gearing up as the government plans to introduce the world's first AI digital textbooks in 2025.

AI digital textbooks are expected to change the paradigm of public education, not just as a supplementary booklet to existing textbooks, but as a solution to realize 'personalized education' according to student level and progress.

While the movement to apply AI to public education is only now in full swing, there is a company that has been waiting and preparing for this change for about 12 years. This is Classting, an edutech startup founded in 2012 when the term AI was still unfamiliar. Classting, which entered the market early, has more than 9 million users in elementary, middle, and high schools in 47 countries around the world with its own AI education SaaS (Software as a Service). With more than 70% of primary digital pioneer schools piloting AI software ahead of the AI digital textbook rollout, and more than half of all digital pioneer schools by 2023, Classting AI is trusted by the education community.


The company turned profitable in the second half of 2023 as pilot schools paid for its AI education SaaS. While the company has been profitable before, this is the first time it has been profitable with a paid subscription model rather than advertising. As such, Classting AI is at the forefront of the public education-intersects-AI movement, benefiting from a paradigm shift in public education and a dramatic allocation of funding from the Department of Education. But it took more than a decade of hard work for the company's seemingly smooth sailing to materialize. The "integrity" of Classting CEO Jo Hyeon-gu, who quit his job as a stable elementary school teacher to develop software for public education despite the widespread belief that "public education is not a business," is finally seeing the light of day in the AI era.

Jo, who started his business with the question, "Why is public education the same when technology is changing every day," has been helping schools digitally transform their classrooms by providing online classroom software for free until 2019, when he launched his first paid service, Classting AI.


By the startup's rapidly expanding clock, growth has been slow. However, since Mistletoe, a private equity firm run by Son Tae-jang, the younger brother of SoftBank Chairman Son Jeong-ui, joined as an early investor, SoftBank Ventures, Atinum Investment, Samsung Venture Investment, and Korea Development Bank have invested more than 30 billion won in the startup and have been waiting for Jo's integrity. What was the blueprint for his dream public education innovation that attracted co-founders, employees, investors, and other stakeholders to the "unprofitable" market, and even moved the hearts of stakeholders in the conservative education world. "I didn't do it because I knew the market would open, I did it because I had to do it," says Jo Hyeon-gu. What are the "tasks to be done" for CEO Jo Hyeon-gu? We spoke with CEO Jo Hyeon-gu, the founder of Classting, and Jang Yu-jin, the head of the business strategy group who has been working on developing the revenue model, about the company's journey in preparing to seize opportunities.


"For schools to change, 'teachers' need to change"
A service that puts teachers before students

‘A service designed solely for the convenience of teachers’

This is the best way to describe Classting's beginnings. When Jo joined Dongbang Elementary School, Incheon in 2009 and taught for four years, he was an enthusiastic first-year teacher who cared about each and every student. But each student was learning at a different pace. With a heterogeneous class, including North Korean defectors and foreign students, it was impossible to achieve the same academic goals for students of different levels and achievements. It was hard for him to keep an eye on 20-30 students per class and 200-300 students in total, and to teach them at their own level. He couldn't gauge whether they were on track or not, so he was forced to give them one-sided instruction.

But it wasn't an environment that allowed teachers to focus solely on preparing for class, either: there was a lot of printing, collating, grading, statistics, and other tasks that needed to be done outside of school, and it was hard to keep up with home communications to parents, signatures, and student assignments. As he struggled on the front lines of public education, he realized that personalized learning, and even mastery learning1 195% or more of all students achieve 90% or more of the learning objectives 닫기 , was not possible in a public education structure where one teacher is responsible for multiple students.

This practical barrier is what prompted him to turn to technology. He decided to borrow the power of technology to help him. "I couldn't physically handle that many students, even if I took breaks to catch up with them after class," Jo said. "I wondered how I could teach them well and efficiently, and decided to use technology to free up time." It was a natural idea for him, who has always been interested in technology, having graduated with a bachelor's and master's degree in computer education from Daegu National University of Education and Seoul National University of Education, respectively.

In doing so, he created a service that would ease the burden on teachers and save them time. In 2011, he launched the app, which combines a basic learning management system(LMS) for sending announcements to students and parents, receiving assignments and signatures, and an educational social networking service(SNS) for two-way conversations. Since the service was built for his own use, he focused on solving the technological needs of teachers and gradually moved everything that happens in the classroom, such as sharing resources, checking homework, and providing career counseling, to mobile and the web.


In order to encourage active communication, he also made sure to respect teachers' privacy, which is one of their biggest concerns. Many teachers were feeling overwhelmed by the number of unanswered questions from parents calling their number. They were also tired of students commenting on their personal social media accounts. He listened carefully to their concerns and made sure to keep their contact information and personal information private so that they could communicate in a timely manner without worrying about their privacy being compromised. He took security seriously by allowing teachers to create online classes and share invitation codes.

He also needed to make it as "user-friendly" as possible to encourage non-tech-savvy teachers to use it. A teacher-friendly user interface (UI) was paramount, as students are quick to pick up technology, but teachers are often intimidated by it. Jo asked the 50-year-old department head teacher of the next class to try the app without any explanation, and then checked where he got stuck or felt uncomfortable. He then revised the UI until the department head teacher was able to master it without any questions. After recommending the initial version of the app to other teachers and promoting it on Facebook, the service began to spread to the next class, the next grade, and neighboring schools as word of mouth spread. Without any marketing, the app gained 100,000 users in the first year, and by the time he decided to start his own business in 2012, it had grown to hundreds of thousands. The cost of the server was over 10 million won, and the burden grew. At first, he survived without starting a business, but he realized that he needed to start a full-fledged business to run a sustainable service and cover the growing costs. That's why he decided to quit his job as a teacher and go into business.

This intense focus on solving teachers' pain points has become Classting's competitive edge. "Just as doctors need to be at the center of changing hospitals rather than patients, teachers need to be at the center of changing schools," says Jo, "and we made it clear that the end customers we were ultimately trying to satisfy with our software were the teachers who teach the students, not the students themselves." As a former teacher himself, he understands the challenges in the field, but to get the most up-to-date feedback, he also involved Classting's EDULAB, a group of 20 early adopter teachers, in the planning stages of the service. In addition, they selected 140 ambassadors who use Classting well to explain real-life use cases to schools that are introducing Classting, and they have an 800-member advisory board to listen to improvements and suggestions.


The mission of 'personalized education'
AI as a solution to save public education

He believed that public education should be able to incorporate the latest technologies that are competitive in the private sector, so he sought out people with strong development capabilities. At the honeymoon of his old friend Yu Jae-sang, the current CTO, who had just finished his master's degree at KAIST and was still working at the government research institute, he bombarded him with 100 pages of business plans, and he began to seek out his KAIST contacts to gather people who shared his vision of changing education. At first, they thought that the more features they could add together, the better the service would be, and they worked hard to update the gamification features to help students self-paced learning. There were times when "fun" became a priority because they wanted to make the platform as engaging as possible for students. But in the process of focusing on what students would be interested in, the company complained that they weren't catering to their end customers: teachers. What they really wanted wasn't just more students using the app; they wanted a "tutor" of sorts, someone who could objectively assess their students' learning levels and tastes and help them improve the quality of their teaching.

As Classting went through the trial and error of figuring out which features to power, it reaffirmed its mission to help its core audience-teachers-enable "personalized education." Ultimately, it decided that an "AI assistant" trained on students' data would be essential to realize its vision of personalized learning, and that it was urgent to penetrate as many schools as possible to collect data for AI training. At the time, the term AI was unfamiliar and there were many doubts about the accuracy and reliability of the models, but Jo was nevertheless convinced that the mission of mastery learning could not be accomplished without AI. "What we do best is combine technology with education to make education more efficient," he says. "In an era where AI is developing rapidly, I think the role of a platform that bridges the gap between people who only know education and AI is bound to grow."


This mission was realized as the "killer feature" of Classting's AI service. A key feature of the service that has been well-received by teachers to date is the AI-based "diagnosis" of student learning. It identifies where a student has gaps in their past learning, i.e., where they lack the foundation for a certain unit in a certain grade. The AI starts with a minimal set of questions and works backwards through the curriculum until it finds where the student is struggling and where they lack understanding or application of concepts. Diagnostics is also a prerequisite for personalized "prescriptions," such as giving extra questions and lectures on unknown concepts like least common denominator and factoring. "Teachers have responded well to the diagnostic feature, calling it a revolutionary feature that can transform education," says Jang.

In order to improve the accuracy of these diagnoses and prescriptions to become an education solution that will help students’ mastery learning, they needed more data from students, and they needed to quickly create a variety of features that teachers would need for every situation that arises in the classroom and convince them to use them, so that they could attract as many students and parents as possible to their service. This is why they had no choice but to continue to distribute the service for free.

Public education is conservative about change, so the linear growth in subscribers - 1 million in 2015, 2 million in 2016, 3 million in 2017, and 4 million in 2018 - didn't translate into immediate revenue. They needed money and time to sustain themselves while they continued to grow their customer base and accumulate data.


It takes money and time to survive
In conservative public education market

(1) Convincing investors to tolerate "slow growth" instead of fast growth

The biggest problem with public education services was that they didn't make money. This was a fatal weakness that jeopardized the existence of the business itself. In the early days of the company, the prize money from various startup competitions, including the Global K START-UP in 2012, became a source of support, but the company had to find a revenue model or attract external investors to survive.

The ideal revenue model Jo envisioned from the beginning was to charge frontline schools for educational SaaS, accumulate student data, and eventually implement AI-powered personalized learning solutions. However, there was no market for public education services in Korea, and the "30 trillion won Korean education market" was all about the private education market. Schools bought hardware for classrooms, such as desks, PCs, and TVs, but they didn't think about paying for edtech solutions from private companies. Since it was common practice to use programs that were outsourced and distributed to schools by the Ministry of Education or its agencies for free, the perception was, "Why pay for something that the government gives away for free? Naturally, private technology companies could not compete, and free software was often left underutilized and replaced when the head of the organization's term expired. As a result, large companies withdrew from education services after two years, and education startups were forced to turn to the private education market and look to parents' wallets.

There was no immediate way to monetize the service, as the only way to compete with free is to respond with an equally free service. Providing a free service requires funding for development and operations to stay afloat, so the first step was to find investors whose expectations aligned with the company's own. The problem was that Classting was not an attractive business for investors looking for a quick return and immediate rewards. Jo considered a social enterprise early on, and while the social value it could provide was clear, the path to monetization was unclear.

In the end, the only answer was to emphasize the authenticity of the company's values. From the beginning, Jo didn't promise lofty goals when meeting with investors. He emphasized that education is a slow-moving, conservative market, and that "waiting" is essential until there is a realization that technology needs to be integrated into the classroom. Rather, he pointed out that the public education market cannot grow as rapidly as the gaming market. He wasn't promoting rapid growth, but rather slow growth. Instead, he made it clear that even if it's a slow process, once it gets off the ground and all schools are covered, it will be a robust and irreplaceable market. This honest acknowledgement of reality, and his sincere commitment to personalized education that builds on students' strengths and capabilities, has filtered out investors who are looking for immediate returns in favor of larger firms that have the stamina and deep pockets to see this through, and who understand the need to transform public education.

(2) Finding short-term revenue models that don't compromise the essence of education

But they couldn't rely solely on outside funding, so Classting devised two business models to realistically generate cash flow. The first was advertising. With millions of users, Classting was well positioned to monetize advertising, just like any other social media. Even though the ad rates were four to five times more expensive than the big portals, ministries of education and school boards responded to the ad products. However, Classting's criteria was that the ad revenue should not take away from the essence of education. While several companies targeting students, such as gaming companies, offered large sums of money, Classting stuck to its philosophy of not advertising anything that could interfere with students' attendance. They avoided advertising products or services that parents might not want to see on an education platform, even if they had a need to buy them. They didn't want to advertise products or services that parents might not want to see on the platform, even if they had a need to buy them.

The second was commerce. It started with the idea of connecting education buyers and sellers so that all education-related expenditures could be made on the platform, and it established itself as an education vertical commerce platform. The service, called "Edustore" by Classting, listed curated books, stationery, school supplies, and other products and handled sales and delivery, earning a commission from the middleman. It also operated an experiential learning marketplace that recruited and connected students to one-day classes, museums, art galleries, etc. for hands-on learning outside of school after school and on weekends. While the company was increasingly relying on commerce as well as advertising to generate revenue, margins and operating income were low compared to the amount of money and time it had to invest. The company decided that it would be a hindrance to focusing on its core business of developing AI learning software, so it transformed itself into a SaaS company, started charging for services, and discontinued its commerce business in 2020.

While these short-term revenue models were being explored, the company continued to invest in its long-term revenue model, SaaS monetization. The Learning Card project, launched in July 2016, is a prime example. The project, which aimed to personalize education for the mobile era, was the beginning of personalized services and the predecessor of Classting's AI service today.

Learning Card is a term that refers to a "standard" for digitizing educational content. The idea behind the project was that in order to personalize education with AI, they need a lot of digitized content, and Classting didn't have any. Since the company couldn't create all the learning materials from scratch, it had to partner with an off-the-shelf company. Most education companies have been in the business of selling paper textbooks, which are content-centric, so the need for technology was not yet recognized. For the incumbents, there was no reason to invest in development capabilities when there was no way to monetize the public education market, even with digital transformation, and no promise of AI in education.

Learning Card was an attempt to fill this gap in education big data and bring scattered offline content together online. Classting curiously held a public meeting and invited textbook publishers such as Chunjae Education, Visang Education, MiraeN, and other content companies such as the BBC and Disney, promising to "breathe life into their content and create new business" and computerize their manually typed content in exchange for access and utilization. It was a no-brainer for the content companies to take the hassle out of the equation, expose them to Classting's user base, and ultimately get paid.

It took a long time and a lot of money to build the big data base using outsourced labor, which is still ongoing to this day. But the Learning Card project, which started as an open beta service, didn't come to fruition immediately. "Pikicast, a content platform, was in vogue at the time, and short-form content that can be consumed quickly and easily on mobile was gaining popularity," says Jang, "So we expected that personalized education would be possible by recommending short-form content that meets individual tastes and needs, but it was premature." Schools still didn't understand the need for AI-powered personalized education or were willing to spend the money, and the lack of a market didn't translate into revenue.



Assets of 'Big Data' and 'Trust'
Opens the opportunity for the Prepared

Although there were no immediate results, the big data accumulated through the Learning Card project laid the foundation for the initial version of Classting AI and its diagnostic capabilities in October 2018. The integration of data through partnerships laid the foundation for business expansion. By accumulating textbook data for each grade level, the AI was able to accurately diagnose where a student was stuck or needed to improve and recommend the right content for the student.

In fact, Classting's strength is that it has all the data related to a person's education. This data is divided into two main categories: social data and learning data. First, social data is collected as unstructured data, such as students' various discussions, collaborations, and comments in the online classroom. Thanks to its function as a social media that enhances two-way communication, it collects spontaneously generated data such as student relationships and emotions. Next, learning data, such as problem solving and submitting assignments given by teachers, is collected as structured data. "On average, about 30 million pieces of data per day are accumulated in the Learning Record Store (LRS), Classting's educational data repository, which is the company's core asset," said Jang. "Since students' activities are accumulated in the app and this data is released in the form of reports, teachers are also encouraged to teach more activity-oriented classes."

Being the only edtech startup in the public education sector since 2012 has also helped Classting build trust and brand reputation in the conservative education community. In his spare time, Jo has participated in consultations with various city and provincial superintendents, ministries of education, and school boards, and spoken at domestic and international conferences organized by the Principals' Association, Korea Education & Research Information Service (KERIS), and others to talk about edtech-based public education innovation. This time spent advocating for the marriage of public education and IT has made Classting's presence felt in the education community.

(1) First Chance: Digital transformation of education sparked by COVID-19

After accumulating data and trust assets and perfecting an early version of Classting AI for personalized learning, Classting was presented with an opportunity. The novel coronavirus (COVID-19) pandemic marked a new inflection point in edtech. When the pandemic hit and virtual learning became the new normal, even schools that were reluctant to make the digital transformation started looking for online class platforms. In this situation, no one could compete with the big data and trust that Classting, which had only sold one well of public education, had built up. Not missing this opportunity, Classting quickly penetrated the market by expanding features such as automatic attendance processing, messenger, and COVID-19 self-diagnosis registration required for remote classes, and offering a one-month free trial service. The biggest change brought about by COVID-19 was the realization that schools could pay for Classting AI, an AI learning service, as edtech budgets were being organized around city and provincial education departments. The public education software market actually opened up for the first time.

In response, Classting began supplying Classting AI to Daegu City Office of Education, and the first revenue from AI education SaaS sales, which the company had envisioned since its inception, came in 2019. As word of mouth spread among teachers and parents in Suseong-gu, Daegu, and other areas with high education enthusiasm that "AI can help students study more effectively," the service spread to neighboring schools. Of course, it was a loss-making business for the company at the time. The market was nascent, and few people had ever spent money on software before, so they priced it cheaply. However, it was a significant first step in terms of launching a paid service.

As the COVID-19 changed the education community's perception of digital transformation, the demand for individualized education gradually surged. Teachers and parents at schools were the first to propose the adoption of Classting AI. In particular, as budgets were allocated to purchase software to improve the academic performance of students with low basic education, schools began to actively spend money, and business agreements were signed with education offices such as the Incheon Metropolitan Office of Education, Gyeonggi Provincial Office of Education, and Seoul Metropolitan Office of Education. As the company accumulated cases of academic improvement among students who fall below basic academic standards, it became easier to convince schools. By April 2020, DAUs (daily active users) had jumped 2.4 times compared to pre-COVID-19, and the number of subscribers had increased to 6.5 million in April, including more than half of all teachers, students, and parents.


(2) Second Chance: The emergence of large-scale language models and generative AI

The advent of large-scale language models (LLMs) such as GPT and LLaMA was another catalyst for Classting AI's growth. The ability to deliver personalized learning content based on a comprehensive analysis of a student's conceptual understanding and performance has been dramatically improved by leveraging the Transformer neural network model, the foundation of super-generative AI. Recently, the research on math problem-solving algorithms was accepted as a main paper at EMNLP, the world's most prestigious natural language processing conference. "The performance of the GPT model fine-tuned using public education big data is far superior to that of other models," said Jang. "It has excellent interpretability and stability even when running large-scale data, and the correct answer prediction rate is as high as 94 percent."

According to the company, the AI engine doesn't show students a set of questions, but instead calculates the correct answer in real time as they solve each question and recommends the next question. The calculation takes a fraction of a second, so students don't even notice that the questions are changing to suit them. Furthermore, GPT allows students to create their own multiple-choice, narrative, and essay questions, and they can also self-check their answers and explanations to make sure they are correct.

Advances in generative AI have also solved a challenge that has plagued the company. Until now, the inherent limitation of AI engines has been their inability to perform well in a "cold-start"2 2no data on the program side. A cold start for a recommendation system is when it fails to make the right recommendations for new users or users for whom you don't have enough data 닫기 environment, where they don't know anything about the student or the curriculum. As a data-driven engine, there was no way to avoid the fact that it would perform poorly when data was scarce. No matter how much they advanced the engine, the AI needed time to get up to speed with the first students, the first schools, the first curriculum and problems it saw. This was a barrier to customer scale. For those buying the service, the prospect of having to wait months for it to work properly was a deterrent.

However, recent in-house research has enabled the integration of multiple AI models, which has dramatically solved the cold-start problem and provided a blueprint for Classting's sales, especially international expansion. Even in an untested market like Saudi Arabia, after just a few pilots, the company was able to predict how well a student would do on a test based solely on social data, such as the everyday vocabulary they use, and suggest personalized learning. Classting expects the advent of generative AI to accelerate growth. For example, it can recommend questions and video lectures that match a student's level and progress, but it can also create content itself. Even before the introduction of AI digital textbooks, there is still a perception that content companies are in a strong position to organize the content of textbooks and outsource the technology, but as generative AI develops and deep-tech competitiveness becomes more important, technology advantage can lead to content advantage, according to the company. "Jello, an educational GPT and AI assistant teacher chatbot that has already been launched, will be able to generate a formatted document when a teacher asks it to write a home letter, and suggest similar questions to the one the student got wrong to encourage further learning," Jang said.



Going global again with AGI
Top-down rather than bottom-up

This isn't Classting's first attempt at international expansion, as the company is poised to go global thanks to advances in generative AI. The founding members of Classting started looking overseas as early as 2015, when they realized that their service was competitive enough. Competing services such as Google Classroom emerged later than Classting, in 2013, and it was also different from Classting, which was born out of a need in the education sector, as it was an educational version of Google's G-Suite, which started as an enterprise platform. As in Korea, the company decided to target frontline schools by partnering with teachers, who are its core users and target audience, and launched a free version in 15 languages in 25 countries, including the United States, China, Japan, Taiwan, and Singapore.

They simply opened a free version of their online classroom in a foreign language without any marketing or promotional activities, and the organic traffic started coming in. Overseas, they saw Korea's high level of education enthusiasm as something to benchmark rather than a social problem, so they were curious about Korean education brands and services. In particular, at the beginning of its overseas expansion, the company had targeted the U.S., Japan, and China, but unexpectedly received users from Taiwan, a country with a similar education system to Korea. Seeing the potential in this market, Classting hired a Taiwanese employee as a manager and began traveling to education fairs across Taiwan, and when schools or education bureaus inquired about how to use the app, they provided training to teachers. One employee was barely enough to keep up with the volume of inquiries, but since it was a free service, they couldn't afford to invest a lot of resources in expanding support. They signed a business agreement with a publisher to acquire Taiwanese educational content and develop a localized product, but the publisher launched an online classroom that copied Classting's UX/UI down to the colors. However, they didn't have a legal contract and lacked the resources to fight the legal battle. Despite this lack of localization, Classting's popularity in Taiwan grew steadily with basic features, and the service spread to more than 1,330 elementary, middle, and high schools, about 30% of all schools, during the peak of the COVID-19 pandemic. However, they still didn't have enough content or data to launch additional features to monetize. The same was true in Japan, where 338 schools signed up in a similar fashion to Taiwan.

In the United States, school districts in San Diego and elsewhere have shown interest. The University of California, Los Angeles (UCLA) even offered to collaborate on a study to see how personalized learning in online classes could improve student performance. In response, the company hired an American employee to conduct joint research and respond to inquiries from the local school board, and Jo personally traveled to the United States to sign an MOU with the school board. There was also a time when they signed an MOU with the Colombian Ministry of Education in South America as part of ODA with the Incheon Metropolitan Office of Education to introduce Classting to schools across Colombia. However, the North and South American regions also lost momentum when the COVID-19 outbreak hit before full-scale product development.

The problem was money. Despite the lack of competition, the company's first attempts to go global quickly and despite the lack of response, the company's first attempt to go global fizzled because everything they tried turned into costs. Setting up local subsidiaries and providing services to schools requires dedicated managers and support staff in each country, and since the service was offered for free, the deficit accumulated with each additional country. Even a universal solution would require resources to acquire local content and package a product that would sell well in each market. Instead, Classting decided to reduce the number of supported languages to six, perfect the AI-powered paid version in Korea, and gain operational experience before re-entering the market.

But the experience wasn't useless. Through trial and error, Classting learned that a bottom-up approach like Taiwan's, with employees walking the streets to frontline schools and teachers, was costly and limited in scalability. Collaborating with universities or partnering with school boards, as in the U.S. and Colombia, was more effective in the public education business, where trust is key. They realized that they needed to quickly penetrate the public education system from the top-down, rather than reaching out on a personal level to build awareness and partner with local content companies and protect their intellectual property. It was also important to find an efficient route, as they could not wait and invest resources to build trust in every country every time, as they did in Korea.

Based on these insights, Classting is looking to expand globally again in 2024, when paid services are in full swing in Korea and competing services are starting to emerge overseas. The company expects this to be different from the first time, as it has solved the problem of cold start, which requires a lot of data when entering a new region, and has advanced the AI technology required for paid services. "We believe it's time to go global again to capture the market ahead of our competitors," said Jo. "There is no AI-based personalized learning solution yet, so we plan to target overseas markets again with premium services."

It's unclear whether Classting's growth will be fueled by market turbulence, such as the introduction of AI digital textbooks and the evolution of GPTs, or whether it will be challenged by increased competition. But the fact that private tech companies are competing in what was once a barren public education market is a sign that the company's mission of advancing public education - one that investors have believed in for more than a decade - is beginning to be realized. "Teachers' role is to help students effectively reach their educational goals, but their expertise has not been realized due to practical and physical limitations," Jo said. "We hope that technology will solve these difficulties, revitalize public education, and attract more talented people to the domain so that their expertise in education can be recognized."

DBR mini box I: Success Factors and Implications ❶ Beyond Public Education, Lifelong Education and Corporate Education Expands Possibilities

The scope and potential of edtech, the convergence of education and technology, is expanding every day. In particular, as face-to-face education has been restricted due to the COVID-19 pandemic, edtech has rapidly penetrated into the education of all generations, including school education, corporate education, and adult education, and AI algorithms have been developing at a rapid pace, making "AI in Education (AIED)" a subfield of educational engineering. One of the most effective applications of AI in edtech is the world's first AI digital textbook, which will be introduced in Korea next year. The edtech industry, including AI digital textbooks, is expected to grow steeply, reaching 10 trillion won in Korea next year, according to the Ministry of SMEs and Startups in 2021. In particular, AI is developing at an unimaginable pace, so it is impossible to predict what form the AIED industry will take among edtechs. What is certain is that the success of AIED depends on serious consideration of the intrinsic value of education, not just technological innovation. In this context, Classting's focus on the intrinsic value of education is a glimpse into the future of AIED. Classting has been pursuing the essence and public nature of education from a long-term perspective to achieve learner-centered and mastery learning, not just converging education and technology. In the process, it has established itself as a leading company in the field of AIED in Korea by securing high-quality education big data over the years to prove the public nature of AIED and the scalability of educational targets. Classting's case shows that it is possible to go beyond ensuring equality of educational opportunities for individual learners to achieving equity of educational outcomes, where all learners can rea¬ch their full potential and achieve mastery learning. This has important implications for the value of education in a rapidly changing world and the responsibilities of education in our society. 1. Securing quality education big data by focusing on the 'longevity of education' The effects of education are not observed in the short term, but only in the long term. Classting's focus on the long-term nature of education led them to establish and implement a strategy to steadily accumulate high-quality educational big data. This strategy has created a comparative advantage and a barrier to entry for the company, as educational big data, a key factor that determines the success of AIED, is not only difficult to secure in a short period of time, but also difficult to ensure the accuracy of AI if the quality is low. Classting's early users were mainly teachers in public education who were accustomed to using free software. As a result, adopting paid software was a challenge, as it entailed a number of difficulties, including cost and adapting to a new system. Nevertheless, in order to secure quality education big data, Classting chose a strategy that emphasized the intrinsic value of "education" and attracted long-term investors by offering its paid services for free for a long time. Classting began accumulating educational data by distributing free educational software with cutting-edge technology, and then focused on advancing its AI technology to maximize educational effectiveness. To implement the Learning Card project, launched in July 2016, Classting accumulated big data by digitizing and standardizing a large amount of educational content from existing educational companies, which led to the launch of the initial version of Classting AI and its diagnostic capabilities in October 2018, two years and three months later. This education big data gave the company the opportunity to ride the wave of digital transformation (DX) that hit the education industry during the COVID-19 pandemic, and the timing was perfect to leverage the accumulated education big data to maximize the accuracy, convenience, and accessibility of AI to enable mastery learning. 2. Developing an AI-based cloud platform that is a "public good” In 2017, economist Barry Eichengreen of the University of California, Berkeley, defined the current era as the "age of hyper-uncertainty," which goes beyond the age of uncertainty. In his view, we have entered an era of extreme uncertainty, surpassing the era of uncertainty described by American economist John Galbraith 40 years ago. In this uncertainty, personalized learning is not an option, but a necessity. In order to optimally respond to uncertainty, learning goals and learning strategies must be tailored to individual characteristics. In this era of hyper-uncertainty, the knowledge and skills needed by learners are not objects specified by instructors, but creations created by learners. In other words, we have reached a situation where we need to consider a new learning theory beyond the paradigm of "instructivism" that aims to effectively convey absolute facts. An alternative learning theory is Constructivism, where the learner constructs, generates, and creates knowledge, and the instructor facilitates this process. To apply this learning theory in the classroom, educators need to identify the factors that enhance or hinder the role shift to facilitator. With this in mind, Classting wanted to bring the cloud, which is based on AI algorithms, to the classroom. The cloud refers to virtual servers and programs that can be accessed using a vast network. A cloud-based LMS has the advantage of enabling teachers to engage in teaching and learning activities independent of time, space, and device, is easily scalable, and can be connected to other cloud services. Classting's CEO brings his own experience as a teacher to technology. For example, he realized that teachers had difficulty managing and communicating with their classes, so he developed an educational app that combines LMS and social media functions, and actively collected feedback from teachers to improve the service. This is similar to the fundamental thinking of AI, which learns from a large amount of data to identify data patterns and come up with optimal solutions. As a result of this pursuit of optimizing individual mastery learning, the cloud system itself is the private property of the company, but the platform is a public good. The fact that Classting AI has had a positive effect on improving underperformance in multicultural classes and math learning for students in rural areas demonstrates its value as a public good and shows the potential for AI-based edtech to address educational inequality in the future. 3. scalability beyond schooling to higher, lifelong, and corporate education The advent of the Transformer neural network model, the basis for super-generative AI, and advances in large-scale language models have solved the "cold start" problem of having no information about learner characteristics and curriculum. These technological advances have particular implications for adult education. Unlike public education, which is characterized by large enrollments and compulsory education, the learners in higher education, continuing education, and corporate training are adults, and the curriculum for them is relatively liberal, unlike public education. In addition, the goal of education is not just to acquire knowledge, but to build practical competencies that can be used in real life and work, such as increasing employability, developing job skills, and improving organizational culture. In this environment, it's important to help individual learners organize their knowledge to meet their needs. However, Classting has been working to overcome this limitation with the latest AI technology and presented its research on AI algorithms for solving educational problems at the world's most prestigious natural language processing conference. These technological advancements could help expand AIED from national to global, and from public education to universities, corporations, and continuing education. While it's unknown if Classting will expand beyond public education in the future, the promise of AI as an effective teaching and learning tool is no longer limited to public education. Advances in super-scale AI models are maximizing the strengths of AIEDs, such as personalized learning paths, instant feedback, and learning analytics, and could easily extend beyond public education into higher, continuing, and corporate education for adult learners. Classting has the potential to apply AI-powered personalized and mastery learning platforms to education across the lifespan. As a leader in the effort to realize the intrinsic value of education - learner-centered and mastery learning - they are expanding the possibilities of AIED. This is a major milestone in creating an educational environment that helps individual learners optimally realize their potential in the age of AI.

Kim Nam-ju Professor of Education, Yonsei University [email protected] Prof. Nam-ju Kim graduated from Yonsei University and received her master's degree in educational engineering from the same graduate school. He received his Ph.D. in Educational Engineering from the University of Utah and was a professor at the University of Miami before joining the faculty of the College of Education at Yonsei University. His research interests include edtech and AI in education.

DBR mini box II: Success factors and implications ❷ "Unbiased and fair" solves a long-standing problem in education

This year, I was lucky enough to attend Bettshow, the UK's education and training technology expo, to get a glimpse of edtech trends around the world. Among the various exhibitor booths, the space I spent the most time in was the Microsoft (MS) session. They showed how AI can pretend to plan lessons, create content, create learning aids, and even design assessments. As I envied their services and thought about how AI could transform public education and free teachers, a Korean company that popped into my head was Classting, a homegrown edtech startup founded in 2012. A well made by a thirsty man, a service irreplaceable Classting is a startup founded by a teacher and has been gaining market attention since its inception. I met CEO Jo Hyeon-gu at various meetings and found him to be a humble but confident man. As a teacher, he knew the reality of public education all too well, and with the support and advice of teachers as users, he was confident in the success of the service. This confidence and assurance must have appealed to investors and stakeholders. But it's the irreplaceable service that has fueled Classting's growth momentum. Classting's services are divided into three main categories: educational social media and LMS, content curation, and AI-powered SaaS. First and foremost, Classting got off to a good start by starting with social media. They successfully launched an educational SNS service that was easy to use, lightweight, and low privacy risk. Existing SNSs were not suitable for use in public education. The nature of SNSs that use mobile phone information to create an account meant that there was a high risk of personal information being exposed. Exposure of teachers', students', and parents' information leads to unnecessary interference, mental conflict, and fatigue. Today, extraordinary measures such as withholding teacher numbers and issuing work phones are taken to prevent unwarranted complaints from parents, but when Classting was founded, even these minimal safeguards were not in place. By allowing teachers and parents to communicate without disclosing their cell phone numbers, Classting has blocked the negative effects of social media services. Meanwhile, the LMS service has helped ease the administrative burden on teachers. Teachers spend a lot of time sending home correspondence, collecting materials, and entering data. In that sense, Classting's service, which partially frees teachers from repetitive administrative tasks, has captured the hearts of the market by removing the "thorns under the fingernails" of users with tweezers. Most importantly, Classting's service is an AI-powered SaaS. Many education companies want to move into the AI business but can't because they don't have the technology and learning data to do so. That's why Classting has been building on this asset for a long time, believing that the combination of personal learning big data and social data can be transformed into powerful information that goes beyond teachers' insights. Teachers may not want to admit it, but AI can diagnose and assess students more unbiased than teachers. The essence of Classting's AI-powered SaaS service is to provide an intelligent platform that outperforms teachers based on accumulated learner experience and teacher wisdom. Cloud-first companies are threatening to enter edtech Classting is a prime example of "AI X education," demonstrating that AI can solve fundamental problems in education. Given the network effect, where value increases as more users are attracted, this effect is expected to multiply as the user base of platform-based businesses expands to frontline schools domestically and globally. Already, Classting has introduced its products to more than half of the digital leading schools ahead of the introduction of AI digital textbooks, and the expansion of this market is expected to be the catalyst for Classting's growth in the future. Mr. Jo Hyeon-gu is a market enlightener and has a long history of trust in the public education market, one of the most difficult to convince. Few startups in the edtech industry survive for more than 10 years, but 12 years in the public education sector is significant in itself. The company has continued to recognize the demand and potential of the market not only in Korea but also in the global market. But now it's time for a challenge. Competitors are emerging with similar services, and "cloud first" companies are offering cross-domain, multi-platform services powered by AI. In particular, the entry of AI leaders such as Microsoft and Google into the edtech market is a major threat. This means that if they don't build a barrier to entry as a first mover in the global market at a faster pace, they could lose their leading position to a quick second mover. To stay in the window of opportunity, they need to be the company that knows the education landscape best, and they need to upgrade their technology as well as strengthen their brand and market loyalty. Being a leader doesn't mean they have to be siloed, and they should look to co-evolve with others in the AI ecosystem. Don't Overlook Data Protection and Evidence As an edtech platform, one of the things they can't afford to overlook is data protection. Collecting and analyzing user learning data inevitably involves personal information, including socially sensitive information such as student academic performance. The legal and moral liability for data breaches and misuse is enough to shake a company to its foundations. A malicious attack on the cloud, human error, or improperly set access permissions that result in unauthorized data access or data disclosure is a huge risk to the company. Especially if it's in the public education sector, it can lead to a lot of public criticism. As a SaaS-based service provider, Classting can only manage this risk by taking privacy and data security seriously. Finally, while many edtech companies claim their services are innovative, they often lack proven results of their effectiveness. Although Classting has already accumulated papers and patents through active research, in the public sphere, such as public education, the standard of decision-making is higher, and securing evidence is paramount to expanding into domestic and global markets. Evidence-based policy (EBP) means that decisions or policies are chosen based on objective evidence, and the most accurate evidence is data. Classting also needs to continue to build on the good research and data behind its AI algorithms. Just as Classting used 800 advisors to identify service improvements in the early days, it will need to analyze the data accumulated on the platform to listen to the needs and desires of users and continue to suggest the best teaching methods to solve the difficulties teachers and students face in the learning process.

Lee Ji-eun Professor, Department of Management Information and AI Business, Hanyang Cyber University [email protected] She majored in educational engineering at Hanyang University and received his master's and doctoral degrees in information technology management from the Graduate School of Information and Communication at the same university. She has worked for Shinsegae, the National Assembly, and is currently the Director of the Education Innovation Center, where he continues to actively apply edtech to content development and learning management.

한달 무료 체험하기 DBR 멤버십 가입하기

인기기사