"I would like to take the training myself."
Hyundai Motor Group Chairman Chung Eui-sun suggested to the management early this year that they take "vibe coding" training, saying this. Vibe coding is a method in which, instead of a developer writing code directly, a generative AI creates executable code when a person explains an idea or feeling (vibe) in natural language, like everyday conversation. Even non-developers can create and use AI agents needed for their work.
Startled working-level staff tried to dissuade him, saying there was no need for the owner to take the training, but he reportedly expressed again that he wanted to participate to properly understand AI technology. Since then, Chung is said to have diligently taken the vibe coding course together with the management.
Jin Eun-sook, Hyundai Motor and Kia's chief information and communications technology (ICT) officer, introduced this anecdote at the "Hyundai Motor Group AX performance briefing" held at the Yangjae headquarters in Seocho-gu, Seoul, on the 12th, saying, "The vibe coding training began with Chairman Chung's strong will."
After taking the training, Chung said, "The first thing management must do is understand the technology," and added, "You have to distinguish what AI can solve and what it can't to inform decision-making." Jin said, "Chung's belief is that unless AI changes the way management works and their receptiveness to technology, it is hard to lead change across the whole organization."
Chung said in 2018, "We will become a company more like an IT company than IT companies." Since then, Hyundai Motor Group has prepared for digital transformation (DX) and AI transformation (AX) since 2019. Starting in 2022, it changed the way it works by adopting JIRA, Dooray, Confluence, and Microsoft M365, and with a "global one data pipeline" it standardized and linked all data in research and development (R&D), production, quality, and sales.
With the foundation in place, last year the in-house Generative AI platform "H Chat Pro" was opened to general employees. In a secure in-house environment, they can choose among external large language models (LLMs) such as ChatGPT, Gemini, and Claude, and the company has also built its own coding system to create work AI agents by linking with in-house data. As of last month, users exceeded 30,000, accounting for about 80% of Hyundai Motor and Kia's general and research staff.
Results appeared on the ground. In research and development, the "crash safety AI assistant" cut the time to search and analyze past crash test data by about 90%. On manufacturing sites, the "AI automated recognition service," which checks whether the vehicle on the line matches the vehicle in the system, was applied to about 70 global production hubs, saving 5.24 billion won annually. The "trolley alignment optimization technology" reduced unnecessary production downtime by about 86%.
In maintenance, AI is being used that immediately presents related data such as repair methods when vehicle errors or symptoms are entered. Linked with global databases, it is said to analyze not only past records but also real-time maintenance requirements worldwide. As a result, response time shortened by 42% compared with before. The service is also operating in the United States, and since June it has been expanded so that more than 6,000 dealers and more than 40,000 technicians in Europe, Asia-Pacific, and Latin America can use it.
There were also growing pains as vibe coding expanded. Jin said, "There was even a case where an AI agent made by a non-developer deleted about 5,000 in-house documents, and it took two nights and three days to restore them." Jin emphasized, "The fact that there were no incidents would mean we didn't actually try it," adding, "Establishing safe usage standards through trial and error is also part of AX."
Hyundai Motor Group is promoting AI infrastructure by dividing it into "upstream (areas where AI is used externally)," which is directly tied to product competitiveness, and "downstream (areas where it is used inside the company)," which supports internal operations. In upstream fields such as Autonomous Driving and physical AI, the plan is to secure proprietary models and foundational technologies. To this end, by 2030 it is building an AI data center (AIDC) in Saemangeum and securing the necessary graphics processing units (GPUs) through Nvidia.
The ultimate goal of pushing AX is vehicle development. By advancing AX, the idea is to translate accumulated data and AI capabilities into product development speed beyond internal work efficiency. Jin said, "We are identifying tasks on how to shorten vehicle development time and how to tackle development with AI," adding, "Vehicle development periods are longer than those of Chinese complete car makers." Jin added, "We have drawn up related plans and are proceeding."
Cost reduction is cited as a task. That is because large sums are invested to use AI. Jin said that whenever reviewing AI projects, the first thing checked is whether it is a problem that needs to be solved with expensive AI or one that can be handled by general algorithms. However, no specific measures to cut costs were presented at the briefing. Jin said, "We plan to allocate resources so that researchers and developers can use them well."