“Many companies expect that adopting artificial intelligence (AI) will lead to immediate operational innovation, but the reality on the ground presents entirely different challenges,” said Yang Young-mo, CEO of Redbrick, while discussing the current atmosphere surrounding corporate AI transformation (AX) projects. The surge in generative AI has led to a rapid increase in the implementation of chatbots, document summarization, and report automation systems. However, unexpected limitations are frequently emerging during actual usage.
Yang noted that it is common to hear that “AI has been implemented, but employees are not using it effectively.” He explained that initial evaluations often focus on the performance and accuracy of AI models. However, in real work environments, operational factors such as integration with existing systems, data management frameworks, and permission issues become significantly more important.
In practice, AI often clashes with existing workflows. One company established a generative AI-based support system, but its utilization fell short of expectations. The system failed to integrate smoothly with ERP (Enterprise Resource Planning), collaboration tools, and internal systems, leading employees to perceive it as an additional program rather than a seamless part of their workflow.
Issues with document management systems are also a recurring problem. Instances arise where the latest documents are mixed with outdated materials, or where different departments have varying management standards, causing AI to generate responses based on obsolete data. Yang emphasized that “in corporate AI, what matters is not just simple accuracy, but also the operational structure that includes the data being used for responses and whether access permissions are properly reflected.”
Yang identified the shift towards a ‘multi-LLM’ (Large Language Model) environment, where companies utilize multiple AI models simultaneously, as a new challenge. Marketing teams may use SaaS-based AI, development teams may opt for open-source models, and customer service teams may employ separate AI solutions, leading to a diverse range of AI applications. However, there remains a lack of systems for integrated management.
He warned that “from a corporate perspective, it may become difficult to track which organization is using what data with which AI,” adding that in a multi-LLM environment, data control, security policies, and cost management must all be considered. This trend is particularly pronounced in industries like finance, manufacturing, and public sectors, where security and operational stability are critical. Recently, there has been an increase in requests for not just AI implementation, but also for data integration structures, permission management systems, and operational governance design.
Yang pointed out one of the biggest misconceptions among companies is approaching AI solely as a technology project. He stated, “If AI is implemented without organizing internal data structures, workflows, and approval systems, problems will inevitably arise during the operational phase. AX is not just about adding AI functions; it is more about redesigning the entire corporate operational structure.”
He concluded, “In the future, corporate AI competitiveness will likely hinge not on who introduces the most advanced models, but on how reliably and consistently AI can be operated within actual workflows.”
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.