Why GS Caltex, Dongwon Group, and Vessel AI Chose Samsung SDS as Their AX Partner

Cases of companies advancing AI transformation (AX) with Samsung SDS were unveiled.

GS Caltex, Dongwon Group, and Vessel AI shared their AX experiences with Samsung SDS at the 'REAL Summit 2026' held on the 8th at COEX in Seoul.

Among them, GS Caltex's case was introduced. Kim Young-ju, head of GS Caltex's VCPO office, said GS Caltex is working with Samsung SDS to reorganize its business processes, data, and systems.

GS Caltex had already introduced AI. However, fragmented processes by department, mismatched data, and disconnected systems and operating methods had become obstacles to expanding AI use.

Accordingly, GS Caltex is moving ahead with reorganizing its business processes, data, and systems in order to lay the groundwork for broader AI use. The reorganization is based on responding to the AI era.

Kim Young-ju, head of GS Caltex's VCPO office, said the company concluded that advance preparation of AI-ready processes, data, and systems was necessary to use AI properly.

GS Caltex has designated ERP as the core system where key processes and data converge, and is redesigning its ERP-centered operating model. Samsung SDS is supporting the redesign of the ERP-centered operating model in the master planning to process innovation phase.

The redesign is being carried out in a direction that standardizes processes that differ by business division and product line from a companywide perspective. At the same time, data standards are being reorganized and intersystem connectivity structures are being revamped.

At the current stage, the scope has expanded to analyzing the core end-to-end processes of AI agents. AI application scenarios are also being identified at this stage.

Kim said the work was not simply an ERP replacement, but a process of transforming the company's operating system so that AI can actually perform business tasks. Park Min-woo, executive vice president at Samsung SDS, said that to incorporate AI into actual work functions, end-to-end process redesign reflecting the optimal roles of people and AI is necessary.

The person in the photo is Park Jong-seong, head of Dongwon Group's DT division.

Dongwon Group selected Samsung SDS as a strategic partner last year to build an AI-based collaboration environment and modernize communication, and introduced the AI-based collaboration solution 'Brity Works.'

Park Jong-seong, head of Dongwon Group's DT division, cited Samsung SDS's accumulated capabilities in related fields, Brity Works's work-AI connection method, and the connectivity of AI-based services as reasons for selecting Samsung SDS.

Dongwon Group is currently using Brity Works's online meeting function, as well as AI-based meeting note organization and automatic distribution features. It is also using real-time translation in meetings with overseas affiliates.

Park Jong-seong said field employees are communicating through Brity Works on mobile, and that communication among field staff is not constrained by time or place. He also said Brity Works serves as Dongwon Group's AI-based communication backbone.

Samsung SDS presented a direction for changing AI's role from answer-centered to actual work execution. Chang Seong-jung, executive vice president at Samsung SDS, said that if AI in the past was focused on searching for answers, the current goal goes beyond question and answer to actually delegating work. He explained that Samsung SDS's goal is also to shift from AI that finds answers to AI that gets work done.

Under this direction, Vessel AI and Samsung SDS built an AI factory based on the 'Samsung Cloud Platform (SCP).' The AI factory is equipped with the ability to provide GPU clusters for large-scale AI model training and inference. The current operational GPU count is more than 8,000.

An Jae-man, CEO of Vessel AI, explained that in a large-scale GPU environment, even a failure in a single GPU can stop the entire training job. He said that if a storage bottleneck occurs, the entire training job can be halted, and if a network bottleneck occurs, the entire training job can also be halted.

Accordingly, the two companies combined their respective strengths to address issues in large-scale GPU environments. Samsung SDS took on the role of providing the data center and AI infrastructure suited for large-scale training and inference. Vessel AI combined its software technologies for failure detection, automatic recovery, and scheduling.

The combination of such infrastructure and software led to efforts to address failures and bottlenecks in large-scale GPU environments. As a result, GPU utilization rose from around 35% on the existing infrastructure to about 90%.

A photo of Vessel AI CEO An Jae-man was also included, and the photo was taken by Yang Seung-gap.

An said Vessel AI is currently using more than 20 clouds, and based on its own experience, it rated Samsung SDS's cloud services as among the best, both domestically and overseas.

Jung Woo-yong, executive vice president at Samsung SDS, said that the conditions for peak GPU performance include a complete setup that also covers high-performance network and storage services.

Samsung SDS then unveiled a new AI function for Brity Works, called 'Cowork.' Cowork goes beyond simple information search and recommendations, and when given a work objective by a user, it plans the necessary tasks on its own. Cowork handles research, report writing, email sending, and schedule registration, and its defining feature is that it carries out the actual work process end to end. Samsung SDS plans to officially commercialize Cowork in October.

Source: IT DAILY · Yang Seung-gap
Original: https://www.itdaily.kr/news/articleView.html?idxno=241474

References

• https://www.itdaily.kr/news/articleView.html?idxno=241474