On the 31st of last month, at the HANDOK Inc.(002390) production plant in Daeso-myeon, Eumseong County, North Chungcheong Province. On the screen displayed by Seo Ui-yeon, Head of Production Support at HANDOK Inc., as he gave his explanation, the plaster factory producing the flagship product “Ketotop” unfolded as a three-dimensional virtual space. From the various pieces of equipment installed on each floor to the product curing room, it was rendered as if the actual factory had been replicated exactly. On one side of the screen, process data—such as equipment operating speeds and dryer temperatures—was displayed in real time. Warning alarms are triggered if a malfunction occurs in the equipment or if parameters fall outside pre-set ranges.
Last year, HANDOK Inc. began implementing a digital twin and an artificial intelligence (AI) autonomous control system at the Ketotop plaster factory through a government support program for building autonomous factories. The goal is to provide a comprehensive, at-a-glance view of the entire production facility within the virtual space and to enable rapid management and control. Moving forward, the company plans to advance to a stage where process conditions can be tested virtually and AI can propose optimal production conditions.
The Ketotop production line operating in tandem with the virtual factoryHANDOK Inc.’s Eumseong plant is a production facility that has been in operation for over 30 years since relocating from Seoul in 1995. Although it is a relatively old plant, it has been undergoing a phased transformation into a state-of-the-art, data- and AI-driven facility through the integration of sensors, computer systems, robots, and solar power equipment throughout the site.
Rather than completely dismantling the existing equipment and building a new factory, the company has maintained the facility’s structural framework while digitally upgrading its production, quality, and energy management systems. A HANDOK Inc. representative on-site described this as “similar to preserving the exterior of a traditional Korean house while modernizing the interior.”
On the Plasta production line, the coating solution—mixed and prepared on the third floor—was fed down to the coater, where the fabric, in roll form, moved at a speed of approximately 13 meters per minute through the coating and drying processes. Afterward, two different coating layers were laminated together, cut to specified sizes, and then moved on to the packaging process, where they were placed into pouches and boxes.
The basic production process—which proceeds in the order of weighing, mixing, coating, curing, cutting, pouching, cartoning, and final packaging—does not differ significantly from the existing method, but AI technology has been integrated into certain stages of the inspection process. AI machine vision cameras installed along part of the packaging line capture images of the products and determine whether they are defective based on trained data.
Director Seo said, “In the early stages of implementation, there were instances of ‘over-inspection,’ where normal products were deemed defective, but we improved inspection accuracy by repeatedly training the system on data from both normal and defective products.” He added, “Since the introduction of AI machine vision inspection systems, over-inspection has decreased, and the yield of the existing packaging process has also improved.” He explained that this has ensured consistent inspection standards throughout the mass production process and reduced the unnecessary rejection of normal products.
Five Years of Data-Driven Development… The AI Factory Is a “Work in Progress”Of course, HANDOK Inc.’s Eumseong Plant is in the process of transitioning to a fully autonomous factory where AI independently judges and controls the entire production process. The digital twin at the Plasta Plant currently serves primarily for process monitoring, while many functions—such as AI-driven autonomous control and quality prediction—are still in the development or validation stages. This is not merely due to technical limitations. It is due to the nature of the pharmaceutical and biotech industries as regulated sectors, where product quality and patient safety are directly at stake.
HANDOK Inc. explains that applying AI to manufacturing processes requires greater caution than in general manufacturing, as it involves proving the reliability of results and documenting the basis for decisions. Director Seo stated, “Because the pharmaceutical and biotech industries are regulated, there are limitations to directly applying AI to manufacturing,” adding, “Even if AI proposes a process optimization plan, a human must assess the risks, prepare supporting documentation, and approve its implementation before it can be applied.”
Technologies like generative AI, where it is difficult to clearly verify the process by which results are generated, are hard to apply directly to pharmaceutical manufacturing. HANDOK Inc. is also conducting proof-of-concept (PoC) tests focused on verifiable technologies such as machine learning and is applying a “human-in-the-loop” structure, where the final decision is made by a human.
Accordingly, HANDOK Inc. has proactively begun the process of data standardization. Since 2019, HANDOK Inc. has been unifying the formats and standards of data generated by production facilities and various systems, and building the infrastructure to collect this data. Since becoming the first in the domestic pharmaceutical industry to introduce an Enterprise Resource Planning (ERP) system in 1997, the company has also integrated a Manufacturing Execution System (MES), a Laboratory Information Management System (LIMS), a Building Management System (BMS), and a Quality Management System (QMS) into this process.
Director Seo explained, “Without standardized data, AI cannot learn properly, and even if it does, significant errors can occur,” adding, “It took five to six years just to build and standardize the data infrastructure.”
As a result, HANDOK Inc. has been selected for five government projects over the past two years, including the Smart Eco-Factory, Autonomous Factory, Manufacturing Innovation through Robotics, Advancement of Manufacturing for Promising Export Pharmaceuticals, and Rapid Commercialization of AI-Applied Products. The total government funding secured through these projects amounts to 5.983 billion won.
A significant portion of this funding will be invested in facilities and systems to be built in the future. HANDOK Inc. plans to introduce six specialized AI agents—covering production, quality, logistics, predictive maintenance of equipment, energy, and safety management—as well as an AI supervisor to oversee them, into the secondary packaging process for solid dosage forms. The company also plans to install autonomous mobile robots (AMRs), automated loading robots, and intelligent CCTV systems that detect fires, collisions, and worker proximity.
The company explains that the goal of automation is less about reducing the workforce and more about redeploying personnel. The plan is to automate repetitive tasks and reassign existing staff to other processes with increasing production volumes or to more advanced roles, such as operations and decision-making.
The Ultimate Goal of the AI Transition: ‘Global CMO’Behind HANDOK Inc.’s accelerated push for manufacturing innovation lies the expansion of its contract manufacturing (CMO) business. The company plans to enhance its production and quality control systems based on data and AI, and to build stable supply capabilities to meet the stringent contract manufacturing standards of global pharmaceutical companies. Last year, HANDOK Inc.’s CMO revenue reached approximately 40 billion won, a 13% increase from the previous year. CMO orders now account for about 40% of Hanok’s total solid dosage form production volume.
It typically takes 2 to 3 years—and in some cases, more than 5 years—to bring a global pharmaceutical company’s product into production. This is because companies must undergo verification not only of their quality systems, production capacity, and cost competitiveness but also of data integrity and compliance with GMP regulations regarding their facilities. It also took a considerable amount of time to transfer Janssen’s products to HANDOK Inc.’s plant after the closure of its domestic production facilities. Accordingly, HANDOK Inc.’s strategy is to proactively upgrade its manufacturing and quality infrastructure to meet the demands of global clients and expand its base for securing new CMO orders.
Director Seo stated, “We are continuously receiving CMO requests, and our current plant has approximately 20% spare capacity,” adding, “By utilizing vacant land within the plant, we can roughly double our production capacity compared to current levels, so we are preparing a phased expansion plan.”
Meanwhile, eco-friendly facilities and environmental management systems are also being upgraded in line with the digital transformation of production processes. In fact, solar panels with a total capacity of 2 MW have been installed on the roof of the Eumseong plant building, as well as in the parking lot and on the bleachers of the sports field. Through this, the plant generates approximately 16% of the electricity it consumes through solar power.
Environmental data is also being managed through a real-time system. Oh Dae-han, Deputy Manager of HANDOK Inc.’s Environmental Safety Support Team, explained, “In the past, we recorded wastewater discharge volumes manually, but through the Smart Eco-Factory Support Project, we’ve digitized the measurement data. Now, we can check in real time via computer how many metric tons are discharged each day.”
Electricity and water consumption, as well as solar power generation, can also be monitored in real time through the system. In the future, the company plans to use an Agent AI to analyze energy consumption by product and production line and detect signs of abnormalities.
