Many global companies claim their business models must become AI-native, but transforming existing models into artificial intelligence (AI)-native ones is unfeasible.
Some firms suggest layering AI developed with Anthropic onto their models, but this is impractical. Heo Jeong-yeol, Workday Korea’s new country manager, addressed these limitations during the Workday Elevate Seoul 2026 press conference at the Westin Parnas Hotel in Seoul on Thursday.
Heo noted that companies using legacy systems like SAP for enterprise resource planning (ERP) and human capital management (HCM) face innovation challenges.
He emphasized that transitioning to AI-native models requires redesigning core structures. Existing models struggle with innovation due to outdated architectures and data structures, Heo explained. To become AI-native, businesses need AI-friendly contexts and architectures with well-designed ontologies. Many firms hit roadblocks in AI transformation because of these hurdles.
Heo pointed out AI’s limitations in critical areas like corporate accounting and payroll, where errors are unacceptable. Effective enterprise AI must balance reasoning with established policies in a unified system, he stated.
He criticized the approach of adding external AI as an afterthought, saying it shifts risks to customers. In contrast, Heo highlighted Workday’s integrated architecture that combines data, logic, and security from the ground up.
Addressing Workday’s low adoption rate in Korea, Heo suggested the company’s capabilities are underestimated locally. He expressed confidence in Workday’s ability to lead AI-driven business transformations.
Josh Zywien, Workday’s Vice President (VP) of Global Solution Marketing, added that while generative AI excels at reasoning and pattern recognition, its probabilistic nature can’t guarantee accuracy – a key concern for businesses.
Workday proposed combining probabilistic AI with deterministic processes to address these issues. This approach aims to balance AI’s flexible reasoning with deterministic guardrails that enforce company policies and compliance, ensuring both accuracy and automation.
Zywien stressed that even advanced AI can be risky if it operates outside company guidelines and authority structures. All AI actions must follow predefined policies and processes. It’s crucial to prevent abnormal AI behavior like unauthorized data access.
