What support is needed for the "AI democratic government" pursued by the Lee Jae-myung administration to take root successfully? The Ministry of the Interior and Safety (MOIS) AI Government Office leading the shift to public AX, the National Information Society Agency (NIA), and private-sector experts gathered in Seoul on the 28th for a policy roundtable.
Private-sector experts, including Kakao Vice President Kim Se-ung, who serves as head of the AI Government Technical Advisory Group, shared changes in corporations' ways of working and their experiences with failure and learning, and offered recommendations on the direction of public AX.
◇ "There must be no disadvantage even if you fail… we need to revise rules so developers do not feel burdened"
Kang Dong-ju, vice president of "IYUNO Global," which supplies subtitles and dubbing to Netflix and Disney, gave the first presentation on the theme of "designing how to work and organize with AI." He said, "The organizational makeup, where non-developer frontline staff are the overwhelming majority, and strong security and contract constraints are conditions similar to the government," adding, "We brought over the agent we used best as-is, but in the new organization, the desired outputs did not come out. What needs to be moved is not the finished tool but the sequence that leads to that tool." He continued, "We must build a structure where reporting failure does not bring disadvantages. The real output of a lab is not a book of success cases but a list of failures," calling for the spread of an organizational culture suited to AI.
Next, Gwak Geun-bong, CEO of 1G Labs and head of the AX Center at Market Kurly, took the mic. 1G Labs, an AI solutions company, was acquired as a wholly owned subsidiary by Market Kurly in June. CEO Gwak presented the AI experiences of non-developer employees and the results of a rapid prototyping approach for test models.
He said, "We stopped trying to persuade in meetings and switched to showing a prebuilt test model," adding, "A single screen of a homegrown prototype became a more accurate means of communication than a 20-page proposal." He went on, "Under the old way of working, it took three months to make the first draft. Now we make three drafts in three days and throw away two of them," adding, "The expense of discarding has decreased, and working time has been greatly reduced."
Kim Jeong-hwan, head of business development at A2D2, a legal tech startup, introduced a case of creating a lawyer qualification verification automation tool using Generative AI, noting, "Check whether an individual inconvenience is a shared team problem, and start by improving team-level processes. That is effective for spreading AI use within an organization."
Attorney Choi Ju-seon of the law firm Minwho presented institutionalization measures to guarantee public officials' direct development for AI use. Proposing a system in which the government bears verification and risk for development programs while copyrights to deliverables belong to the state, Choi emphasized, "It is important to design so developers do not feel burdened by rules." Choi said, "The moment you develop, public officials face two choices in a structure where ▲ informatization project procedures ▲ security reviews ▲ and audit risks all hit at once: develop in secret or give up," pointing out, "If rules cannot keep up with speed, innovation happens outside the rules."
◇ "There is also a wide gap in AI use at corporations… we need ways to prevent duplicate AI expense outlays"
The ensuing discussion focused on practical problems corporations experienced during the AX process and ways to overcome them.
Director General Kim Se-ung said, "Even when a company actively pushes an AX transition, about 40% of organizational members still show reluctance to use AI agents," adding, "For these members, it is necessary to support intensive training for two nights and three days, or a week if needed, to shift perceptions and ways of working with AI." He continued, "Conversely, about 10% of members use AI coding tools very actively, including cases where they use tokens worth $10,000 a month," explaining, "There is a very large spread in the level and manner of AI use within organizations."
He also said, "Similar projects or technology development are being pursued separately in multiple organizations, leading to cases where AI use expenses, including token expenses, are duplicated," and stressed, "Along with spreading AX, establishing organization-level utilization systems and expense management measures is also a key task."
As a solution, Director General Kim and CEO Gwak proposed building a "skill store." A skill refers to a module that standardizes a specific process so it can be used repeatedly. The aim is to first prompt searches for related skills before developing a specific program, reducing time and expense.
Yu Seobong, head of the Public AX Consulting Subcommittee, proposed specializing the coding of an "independent artificial intelligence (AI) foundation model" (Dokpamo). For reasons such as security, when use of external commercial models is restricted, he said a high-performance AI code model for internal networks should be introduced.
For public AX, there was also a suggestion to shift the government document system from being centered on "Hangul Word Processor" to an AI-friendly, cloud-based collaborative environment such as Notion. Calls also emerged for leaders to change their mindset. Yu said, "Training for CEOs and executives is key to spreading AI within corporations," adding, "It is also a task to overcome gaps in utilization within organizations and siloing between departments when introducing AI tools."
Hwang Gyu-cheol, director general of the Ministry of the Interior and Safety (MOIS) AI Government Office, said, "If we tell public officials to use AI yet shift the responsibility for post-development security, verification, and maintenance onto individuals, innovation cannot spread," adding, "Frontline public officials should focus on identifying work problems and experimenting boldly, while the government not only provides a safe development environment and training but also builds a system that takes responsibility for the operation and maintenance of verified deliverables and their spread across the government."