[Edaily Reporter Lee So-Hyun ] SKTelecom(017670)is applying its proprietary artificial intelligence (AI) foundation model, developed in-house, to steel and automotive parts manufacturing sites.

of

SKTelecom’s Proprietary AI Foundation Model First Applied in the Manufacturing Industry

SKTelecom announced on the 25th that it has signed a strategic business agreement with steel manufacturer KG DONGBUSTEEL and auto parts manufacturer Conec to conduct field demonstrations of “manufacturing-specific Agent AI” based on its proprietary AI foundation model. This collaboration marks the first instance of SKTelecom applying its proprietary AI foundation model to the manufacturing sector.

Since last April, SKTelecom has been collecting manufacturing data—including historical reports on process errors and accidents, equipment manuals, and logs—from KG DONGBUSTEEL and Conec, and has developed a demo version of the manufacturing-specific Agent AI based on its proprietary AI foundation model, “A.X K1.”

A.X K1 is a super-large language model with 519 billion parameters. It is designed to handle complex tasks while activating only approximately 33 billion parameters during the inference process. The company explains that although the overall model is large, it can be used efficiently in industrial settings by activating only the necessary components.

SKTelecom, KG DONGBUSTEEL, and CONEC will conduct field trials in the second half of this year by applying the demo version of the manufacturing-specific Agent AI to actual production processes. KG DONGBUSTEEL plans to apply the Agent AI to the cold rolling line at its Dangjin plant, which produces galvanized steel sheets, while CONEC plans to apply it to its casting and machining processes.

KG DONGBUSTEEL and Conec will share manufacturing process data with SKTelecom in real time, and SKTelecom will use this data, along with on-site feedback gathered during the pilot, to improve the performance and inference speed of the manufacturing-specific AI agents. They also plan to expand the agents’ functionality.

The manufacturing site data collected during the pilot will also be used to train the A.X K2 model, which is currently under development. Upon completion of the pilot, the companies will evaluate the commercialization and adoption of the manufacturing-specific Agent AI. If necessary, they will also consider replacing the model with a subsequent series of SKTelecom’s proprietary AI foundation models.

Manufacturing has long been considered a sector where AI adoption is challenging. This is because the digitization of manufacturing site data has been slow, and accumulated data is often generated and managed separately by process and department, making it difficult to utilize AI effectively. Another limitation is that work methods vary depending on a worker’s skill level and experience. A prime example is the phenomenon of “knowledge silos,” where core know-how remains confined to specific skilled workers. In such cases, a site’s competitiveness can weaken when veteran workers retire or leave the company.

SKTelecom expects that by digitizing scattered manufacturing data and the experiential knowledge of skilled workers—and deploying AI agents trained on this data to the factory floor—it will be possible to respond quickly to errors that occur during the manufacturing process. The company explains that this will reduce response times and improve process efficiency.

Security is also a key factor in the adoption of AI in manufacturing. Since process-specific security is critical on manufacturing floors, it is difficult to adopt cloud-based AI solutions that transfer data externally. SKTelecom explained that its proprietary AI foundation model supports not only cloud-based systems but also closed on-premises environments, allowing manufacturing process data to be utilized within the company’s internal environment without being transferred externally.

From Defense to Manufacturing… Accelerating the Expansion of Proprietary AI Foundation Models

SKTelecom has recently been expanding the scope of application for its proprietary AI foundation model. Last month, the company signed a memorandum of understanding with the Ministry of National Defense and the Ministry of Science and ICT to utilize its proprietary AI foundation model in the defense sector. This collaboration aims to apply the model in the defense sector, where security and data sovereignty are critical.

Starting with defense and manufacturing, SKTelecom plans to expand the application of its proprietary AI foundation model to various industries, including finance, public sector, and healthcare. The company’s vision is to combine the data held by each industry with its proprietary model to create industry-specific AI solutions and support the AI transformation of the domestic industrial sector as a whole.

Jeong Seok-geun, Head of SKTelecom’s AI CIC, stated, “For manufacturing sites where security is critical, our proprietary AI foundation model—which allows data to be utilized without being sent outside the organization—is an effective solution.” He added, “Starting with our collaboration with KG DONGBUSTEEL and CONEX, we will accelerate the AI transformation in the manufacturing sector and expand the use cases for our proprietary AI foundation model.”

Bae Seon-woo, Head of the KG DONGBUSTEEL Technology Research Institute, emphasized, “This collaboration has laid the groundwork for introducing AI based on on-site data,” adding, “We will further enhance our manufacturing competitiveness.”

Lee Gwang-pyo, CEO of Conec, said, “Responding quickly to recurring quality issues on the shop floor has long been a challenge for the manufacturing industry,” and added, “We will use AI to improve operational efficiency on the manufacturing floor.”