The Faster AX Advances, the More Organizational Structure Matters: "Side Effects of AI-Generated Errors and Mistakes"

AI Hallucinations Lead to 'Knowledge Decay' Across Organizations
Building Robust Management Systems Matters More Than Simply Scaling Up AI Adoption

While the adoption of artificial intelligence transformation (AX) by companies is increasing work productivity, it also raises the risk of errors and mistakes in decision-making processes. Therefore, there is a growing recommendation for establishing systematic AX organizations that can correct such issues.

On June 22, the Harvard Business Review (HBR), in its report "Don’t Let AI’s Bad Output Ruin Your Organization," pointed out that companies actively implementing AI are experiencing a phenomenon called "knowledge decay." Knowledge decay refers to the decline in accuracy and reliability of information used for corporate decision-making as low-quality AI-generated outputs accumulate within organizations. Just as individuals may experience hallucinations when using generative AI, knowledge decay is a structural problem that arises as poor-quality results accumulate across an entire organization.

AI can now easily produce documents in formats similar to those that previously required human expertise, such as reports and analyses, making it difficult to assess how much human review and judgment went into the final output. There are also concerns that if original materials are repeatedly reprocessed by various AI systems, the initial meaning and factual accuracy may become blurred, negatively impacting not only the company's work processes but also the quality of its AI models. If time is not spent verifying and correcting the facts and context in AI-generated outputs, the productivity gains anticipated from AI adoption may ultimately diminish.

A major limitation is that domestic companies do not yet have organizational systems in place to maximize efficiency after adopting AI. According to Microsoft's "2026 Work Trend Index," only 7% of Korean respondents believe that organizational use of AI will lead to rewards, which is just half the global average. Additionally, only 16% of respondents said their company’s AI strategy has clear direction, compared to the global average of 26%.

As a result, the ability of organizations to systematically verify their AI adoption is emerging as a key challenge for businesses. Experts advise that as AX becomes more widespread, it is more important to build organizational management systems than to simply scale up AI adoption. There is a need to verify whether AI utilization is actually enhancing the quality of organizational decision-making and work processes. Jaeseong Lee, Professor of Artificial Intelligence at Chung-Ang University, stated, "Over-reliance on AI-generated outputs can lead to errors affecting the decision-making stage," adding, "It is necessary to strengthen work expertise within companies." He further noted, "It is becoming the norm to establish work execution guidelines and to separate AI tasks from human tasks."

Yongjin Kim, Professor of Business Administration at Sogang University, also commented, "AI-induced hallucinations are unavoidable, making it difficult to trust and utilize outputs blindly." He explained, "By building ontology data that structures human knowledge, the burden of verification can be reduced. It is also essential to clarify guidelines and responsibilities for dividing work between AI and humans, and to establish an organizational structure that enables effective monitoring."

This content was produced with the assistance of AI translation services.