Innovating the Way AI Works: Samsung SDS’s Stories of Solving Enterprise Problems

As Generative AI and Agentic AI rapidly reshape how enterprises work and solve complex problems, Taehee Lee, Executive Vice President at Samsung SDS, introduced concrete examples of how AI is actually being applied in real-world enterprise settings. Drawing on hands-on experience across data analytics, model optimization, AI-powered software development, and cybersecurity, VP Lee offered a vivid account of the stages an enterprise actually moves through on its path toward AI transformation.

On May 19, the National AI Research Lab (NAIRL) and the KAIST Kim Jaechul Graduate School of AI co-hosted a Distinguished Scholar Seminar featuring VP Lee at the Seoul AI Hub. Held under the theme “Innovating the Way AI Works: Samsung SDS’s Stories of Solving Enterprise Problems,” the event drew around 280 participants in total, with approximately 120 attending on-site and 160 joining online. Researchers, graduate students, and industry professionals showed keen interest in hearing field-tested experience directly from one of the country’s leading IT service companies.

In his lecture, VP Lee emphasized that AI should be understood not merely as something a company “applies,” but as a process of “transforming” the company itself onto an AI-native foundation. Noting that moving any organization from one state to another is inherently difficult, he laid out three guiding principles for leading such a transformation successfully: managing scalability, flexibility, and complexity. The key, he explained, is to build each change so that it can keep connecting to and extending an ever-evolving landscape of solutions and components, while steadily reducing the complexity that lies hidden across enterprise systems.

Building on this perspective, the talk unfolded across four main pillars. These were knowledge retrieval (RAG) capable of accurately handling complex documents that mix charts, tables, and images; efficient training of domain-specialized models that deliver strong performance even under cost and security constraints; AI-agent-driven modernization of legacy financial code; and guardrails and red teaming for the safe use of AI. Walking calmly through the real-world challenges encountered in each area and the research journey behind solving them, VP Lee made clear that the work does not end with securing a single good model. What matters, he stressed, is the continuous process of accumulating and refining the underlying data and evaluation frameworks.

A passage that especially resonated with the audience was his insight into the question of why AI has spread fastest within software engineering. VP Lee attributed this to decades of paradigm evolution, from modularization and object-oriented design to test-driven development and automation pipelines, all of which had quietly laid the groundwork for adopting AI. He suggested that by examining how each industry domain has built up its own working paradigm, one can gauge how ready that domain is for AI adoption. Equally striking was his open stance in releasing Samsung SDS’s own datasets and models on public platforms, sharing not only the solutions but the problems themselves with the broader research community.

Through the lively Q&A session that followed, the seminar became a valuable exchange where the living experience of industry met the perspective of research and education. We extend our sincere gratitude to one of Korea’s leading IT service companies for so generously sharing its practical, hard-won expertise, including the lessons learned along the way. NAIRL plans to continue connecting industry and academia through ongoing exchanges, taking the lead in fostering an ecosystem where domestic researchers can drive both scientific discovery and industrial innovation on the foundation of cutting-edge AI technologies.