What is the reason for this poor performance in non-IT sectors, which are directly linked to a company's revenue growth? The biggest cause is security issues. For most companies, especially large corporations, the biggest obstacle to adoption is the need to export internal company data to external clouds in order to use generative AI models. In fact, in July, there was a shocking incident in the United States where a hacker attack on Snowflake cloud resulted in the leak of call and message records of 110 million AT&T customers. For these reasons, many companies are blocking the use of external LLMs and are attempting to build their own LLMs internally.

In cases where companies try to build their own LLM based applications internally, they can solve the problem of external leakage of internal data. However, another problem arises: the issue of cost and time. These problems occur because it needs to be built directly in an SI (System Integration). Especially if they try to build a platform for data analysis that can utilize LLM as well, the cost and time entry barriers will become even greater. Realistically, this becomes the reason why it's difficult for small and medium-sized enterprises to adopt generative AI, except for large corporations that have the capacity to invest. If they try to solve this by using public cloud services, the problem of data leakage occurs again.