Generative artificial intelligence (AI) startup Wrtn Technologies (Wrtn) has entered the AX (AI transformation) business. It established the in-house independent company (CIC) "Wrtn AX" on Sept. 29. CEO Park Min-jun (30), who served as head of planning and strategy at Wrtn, is leading Wrtn AX.
AX means introducing AI technology into work systems or tasks. The core of the AX business is the "AI agent." An AI agent is an autonomous AI system that makes independent decisions when performing specific tasks. Wrtn AX is pursuing a business that builds customized agentic AI for public institutions and corporations.
CEO Park is one of the six founding members who launched Wrtn, including CEO Lee Se-young. Park graduated from the physics department at Seoul National University and also studied computer engineering and financial economics. While preparing a youth academic conference during college, the members saw the event canceled due to COVID-19 and were saddled with nearly 100 million won in debt. The team's average age at the time was 25. To repay the debt, members took on tutoring and online lectures. They then shifted the conference online, achieved results, and decided to start a company with the remaining funds.
Wrtn, which started with six people in 2021, has now grown to around 110. Its cumulative investment has reached 130 billion won. Recently, it raised its profile with an ad featuring "G-Dragon" as a model. In just four years since its founding, it has secured 7 million monthly active users (MAU) and is now eyeing 10 million.
Park worked as a researcher at the global hedge fund WorldQuant while founding the company, handling AI algorithms. After returning to Wrtn, Park was in charge of the company's new business strategy. The chatbot service and the company's first agentic AI service are also Park's work.
Wrtn, as a startup, established a CIC because the AX business is in the B2B (business-to-business) domain. Wrtn grew with a focus on B2C (business-to-consumer). However, the AX business, which deals with corporations and public institutions, requires an independent decision-making structure because the pace of AI adoption and customization needs vary by client.
Meeting at Wrtn's office near Gangnam Station in Seocho-gu, Seoul, on Sept. 30, Park said, "B2C and B2B have different business DNA. Given the fast-changing nature of the industry, we judged it would be more efficient to operate the internal culture and decision-making steps independently from the existing Wrtn."
Park said AX has four stages. The first is introducing a basic conversational interface (chatbot) to create an environment where employees can interact with AI. The second is adding information to existing data to generate better answers.
The third is using tool-using agents. The AI executes code, accesses internal systems, and operates computers. The final stage is a human-like agent, in which AI learns and adapts like a person.
Wrtn AX is currently building systems to implement the fourth stage. Park said, "We have already built many basic chatbots in stage one through the Wrtn service, and we also have experience building stage-two systems for internal and external use. We are currently running a separate team preparing for stage three." Park added, "We are also building an internal system for stage four." The following is a Q&A with Park.
—I'm curious about Wrtn's founding story.
"In 2019, team members who worked together running the academic conference secretariat founded Wrtn. The structure was to prepay venue fees and recoup them with participation fees. Then COVID-19 hit. Among the participants was a student coming from Wuhan, China. We went to the airport and blocked the student's entry, but we couldn't stop those transiting through China. We ended up canceling the conference and couldn't collect participation fees. We incurred nearly 100 million won in debt at the time. We paid it back by tutoring over Zoom. We moved the next conference online, and after successfully holding it, we not only repaid the debt but also had funds left to start the company."
—What is Wrtn AX's competitive edge?
"There are three. First, Wrtn has operated a B2C chatbot service and handled millions of traffic events. Through various user questions and situations, we know well where agents fail and how to improve them. Second is our UX (user experience) capability. We have been obsessed with making AI that is easy for users to use. The last is our consulting capability. We don't just provide models; we analyze corporations' workflows and diagnose potential bottlenecks, then build customized agents."
—Why do you cite AX consulting as a strength?
"AX is expensive. Initial costs such as graphics processing units (GPUs) and labor are the reason. Each corporation has different bottlenecks. If you don't find AI application points that fit the business process, you won't get ROI (return on investment)."
—Are you seeing results?
"We are carrying out projects with the Culture Ministry, the Gyeonggi Provincial Office of Education, and Amorepacific, among others. Many requests ask us to optimize document work such as marketing copy and press release drafting. We are building references mainly with public institutions and large corporations, and we have reached double digits in consulting requests. With current staffing, it's hard to meet demand, so we are hiring."
—What government support do you need as a growing company?
"GPU support is urgent. The AX business requires large-scale computing resources, but GPU prices run into the hundreds of millions of won, which is a burden. It would be good to set up a separate GPU pool dedicated to AX."
—What are your future plans?
"First, our goal is to generate 10 billion won in revenue by next year. We are also preparing a paid AX lecture program for executives. We are preparing AX businesses in education, such as an "AI tutor" service for students. In the long term, our goal is to commercialize a "human-like agent" that learns and adapts like a person and distribute it to corporations and institutions nationwide by 2030. We could reach the stage of technical completion as early as 2027."