B, the chief technology officer (CTO) of a Korean company A, used 10 million won worth of tokens alone over the past month. The entire team under B consumed 60 million won worth of tokens during the same period. Token costs ballooned as employees developed various artificial intelligence (AI) agents and ran them around the clock.
As AI transformation (AX) becomes an essential survival strategy for companies, Korean firms are beginning to pay attention to tokens, the basic unit of AI computation. While encouraging more employees to use tokens and apply AI to their work, companies are also voicing concerns about the resulting token costs. Industry observers note that efficient token management is rapidly emerging as a critical component of corporate AI governance.
According to the industry on Wednesday, Korean companies such as Upstage and Wrtn are focusing on having employees actively use AI in their work. As the Korean market is still in the early stages of pursuing AX, companies view the spread of AI experience as more important than immediate token management. "There are cases in Korea where 100 million won worth of tokens has been used," an AI industry official said. "We are looking at the share of employees who use tokens to the maximum, but the numbers are not yet high."
However, a growing number of companies are proactively exploring token management measures. LG CNS recently developed and began operating a dashboard that displays employees' token usage. The system takes token usage data from OpenAI as of two days prior and reprocesses it based on the company's own criteria. In addition to the dashboard, LG CNS is working to establish guidelines to help employees use tokens efficiently and boost work productivity.
Data tech firm BigValue has gone further by developing an AI agent dedicated to token management. BigValue previously launched "Bokdeokbang Gajae" last month, a service in which AI agents assigned specific roles converse with one another to provide answers to customer inquiries about real estate. As conversations between AI agents increased in actual operations, token usage surged. In response, the company appointed an AI general-affairs agent to manage token costs. The general-affairs agent tallies the previous day's token usage daily and prepares a cost report.
AI startup Liner allows its developers to obtain light large language model (LLM) API keys to use AI. Under this approach, developers can use tokens regardless of the LLM type. The move is designed to prevent service development delays caused by token shortages on specific models.
Companies are focusing on token management because token usage has recently surged as even employees who cannot code create AI agents and run them 24/7. On top of this, Anthropic, OpenAI and Google have been rolling out pay-per-use billing based on token consumption, increasing the cost burden. Token optimization strategies have become a variable that will determine the success or failure of AX. "We allocate tokens by team, but if a specific employee consumes a lot of tokens on low-productivity tasks, more productive employees end up facing token shortages," an AI startup official said. "We have not yet placed any particular restrictions, but concerns about token management are bound to grow."
Choi Kyoung-jin, a law professor at Gachon University and head of the Korea Artificial Intelligence Law Association, suggested that "developers who already have a high understanding of AI and use it well tend to consume excessive tokens." He added, "It is necessary to differentiate employees who have not used AI much until now from developers, so that they can use AI extensively and go through trial and error."
