The discussion pointed to rapid vertical AI-driven transformation across energy, manufacturing, healthcare, and law, and stressed stronger collaboration between industry and academia to build a competitive domestic AI ecosystem.

Effective vertical AI hinges on three elements: a high-performance reasoning model, agent-based workflows that reflect real operational logic, and proprietary domain knowledge, with emphasis on planning, retrieval, tool use, verification, and multi-step execution.

Concerns about AI sovereignty were raised, noting centralization of frontier models among a few firms, risks of dependence on foreign models, higher costs for non-English tasks due to token inefficiencies, and the need for a fully controllable, high-performance Korean foundation model to boost national competitiveness.

Harvey, a legal AI company built on OpenAI models, demonstrated that competitive value comes from tens of thousands of custom workflows encoding practical legal knowledge, showing platform value lies in mapping and automating industry-specific knowledge structures.

New AI agent architectures were discussed, including a policy LLM trained to govern tool use, planning, and retrieval independent of response generation, plus the creation of simulation environments and synthetic data to advance agent development.

Vertical AI is defined as redesigning organizational workflows around AI capabilities rather than relying on a single model, focusing on embedding AI throughout processes across industries.

At a May SAIX Peers brown-bag at Sogang University, Professor Du-Seong Chang presented a framework for Vertical AI Transformation and discussed its implications for industry change and national AI competitiveness.