알림 목록

Analysis Area:  Ulsan Metropolitan City

Key institutions:  Ulsan Metropolitan City Headquarters, business offices, affiliated organizations, 5 districts and counties

Agenda: AI Declaration and the Transformation Level of Internal Administration, Decision-Making, and Performance

Golden Time Type: External AI–Internal Administration Gap

Reference Date: 2026.08.28

Version: Regional AX Golden Time Intelligence v3.2
 

Ulsan City has established a structure dedicated to industrial AI and regional AX by 2026, featuring the AI ​​Innovation Industry Office, AX Policy Bureau, and AI Industry Strategy Division. Although AI has been elevated from a technology project under the IT department to an industrial and policy agenda directly under the Mayor, the percentage of processes within the main office that have been redesigned for AI remains undisclosed. If only the dedicated organization is expanded while the operational methods of general departments are maintained between 2026 and 2028, the concepts of an "AI City" and "AI Administration" will be separated. While  Ulsan City has established an organization to promote AI, there is a lack of evidence that the entire administrative organization has been AXized.

Ulsan City has signed contracts worth 14.85 million won for the construction of an AI administrative assistant platform and 9.2 million won for service subscriptions by 2026, and has also allocated a related budget. Although civil servants' use of AI has shifted from individual free tools to platforms purchased and managed by the organization, the number of users, repeat usage rates, time savings, and error rates are not disclosed. If the system remains limited to document summarization, searching, and drafting for two to three years, the administrative assistant will merely speed up work but fail to change the policy-making structure. While  Ulsan's administrative AI has entered the official adoption phase, it is not yet classified as being in the task redesign phase.

Ulsan City has implemented an intelligent system that applies AI to tax source discovery investigations in 2026 to reduce the time required for manual investigations and enhance tax equity. While the application of AI has expanded from general document processing to detection and judgment tasks within tax administration, figures regarding additional tax collection amounts, detection accuracy, false positives, and reductions in processing times have not yet been fully disclosed. Unless model performance data for 2026–2028 is verified, it is difficult to assess both administrative efficiency and the potential for infringement on citizens' rights.  Implementation cases for Tax AX have been confirmed, but the verification of performance and accountability is still in its early stages.

Ulsan’s AI spatial information convergence search service invested 1 billion won from December 2023 to July 2024 to visualize generative AI responses within digital twin spatial information. Although spatial administration has shifted from specialized system inquiries to natural language queries and visual policy judgments, the actual number of user departments, the number of policy decisions reflected, and error correction history are not disclosed. If the platform is operated primarily through demonstrations and idea contests for two to three years, it will remain an information search service rather than an administrative decision-making system. While  Ulsan’s spatial administration AI has been established, the spread of continuous decision-making across departments remains unconfirmed.

With the establishment of the AX Policy Officer and the AI ​​Industry Strategy Division under the AI ​​Innovation Industry Office, the departments responsible for AI policy have become clear. However, if the organizational structure is formed around a dedicated AI department, general departments such as welfare, transportation, safety, environment, and taxation may perceive AI as a project of support departments rather than their own internal duties, while the departmental AX responsibility framework remains undisclosed. If the dedicated department and operational departments remain separated for two to three years, the number of AI projects may increase, but on-site processes will not change.  The establishment of a dedicated organization is evidence of driving force, but it is not evidence of AX across all departments.

The Smart City Division is responsible for data-based administration evaluation, public data, the Big Data Center, the Data Hub, digital twins, and spatial information. Although the AI ​​and data organizations are separated and operate in parallel, a shared responsibility structure connecting data discovery, model development, policy implementation, and performance evaluation has not been confirmed. If the organizational boundaries between data and AI are maintained until 2028, even if models exist, training data updates and policy feedback will be cut off. While  Ulsan's organizational readiness is high, the operational integration of data and AI remains unconfirmed.

The AI ​​administrative assistant has adopted generative AI as an official work tool. While the automation of document creation, searching, and summarization has become possible, it is not disclosed whether approval stages, report volume, the number of meetings, or duplicate data entry have decreased. If AI drafts are simply added on top of existing procedures over the next two to three years, the volume of deliverables increases, but the burden of review grows along with it. Although  Ulsan's administration has introduced the tool, there is no evidence that it has streamlined the workflow.

The application of AI to tax audits is a case that transforms the experience and manual comparison of personnel into data-driven risk detection. However, while the reduction in overall processing time is limited if the procedures of on-site verification, explanation, approval, and disposition remain the same after AI detection, changes at each stage have not been disclosed. If only some detection tasks are automated by 2028, bottlenecks will shift to the next manual stage.  Ulsan Administrative AX has been confirmed in the automation of unit tasks, but remains unconfirmed in end-to-end process transitions.

Ulsan City operates public data provision, a Big Data Center, a transportation Big Data Platform, the Ulsan-style Data Dam, and a system for monitoring the status of data-based administration. While the data infrastructure has expanded from a focus on collection and openness to analysis and policy support, instances where the predicted values ​​and subsequent outcomes of major policies have been verified using the same data are limited. If data is used as a basis for evaluation responses rather than for project planning for two to three years, the decision-making structure remains centered on experience and reporting. Although  Ulsan possesses the foundation for data administration, it lacks evidence where data has been used to decide on the discontinuation of budgets or projects.

In the 2026 joint evaluation of local governments, Ulsan ranked first in the achievement of quantitative indicator targets at the city level. While administrative execution capabilities were confirmed to be among the top in the nation, the number of excellent qualitative evaluation cases was not as prominent as that of Busan, Gwangju, Daegu, or Incheon. If an administration focused on achieving targets is maintained for two to three years, the execution power of existing indicators may be high, but the capacity to discover and predict new problems becomes limited.  Ulsan's administration is strong in target execution, but its data-driven problem-finding capability still requires separate verification.

Ulsan establishes and operates a digital twin, 3D spatial information, an integrated database of underground facilities, and a carbon neutrality policy support platform. Although urban information has shifted from 2D drawings and departmental ledgers to an integrated spatial model, the actual number of changes in decision-making regarding urban planning, disaster, transportation, and carbon policies is not disclosed. If utilization performance for 2026–2028 is measured solely by view counts and participation in contests, the policy simulation function remains unverified.  The Ulsan Digital Twin exists as a technology platform, and its role as a policy judgment platform is only partially confirmed.

The integration of underground facility databases and spatial information links location data for various facilities, such as roads, water and sewage systems, and gas. Even if data integration proceeds, if the field update cycle and the locational discrepancies of actual facilities are not managed, AI analysis merely reproduces outdated information with precision, while the recency and error rates are not disclosed. As construction on aging infrastructure accumulates over two to three years, the discrepancy between the data and the actual site widens.  The administrative value of digital twins is judged by data recency rather than the level of visualization, and Ulsan's public verification is insufficient.

Ulsan operates a youth policy platform, a job portal, a smart map, and sector-specific online services. While citizen interaction has shifted from visiting institutions to multiple digital channels, the level of integrated recommendation and processing of services from various departments based on individual circumstances has not been confirmed. If the number of platforms merely increases over the next two to three years, citizens may find it easier to access information, but they will still have to search for and evaluate policies that suit them.  Ulsan's citizen administration has gone online, but the personalized integrated AX system remains incomplete.

Although generative AI-based administrative assistants have been introduced to internal public officials, open AI consultation regarding citizen complaints and tracking of processing results have not been confirmed. While internal productivity may improve, there is no separate evidence to confirm whether the processing times, repetitive paperwork, and departmental transfers experienced by citizens have decreased. If internal tools and citizen services are separated by 2028, the benefits of Administrative AX will be concentrated within the organization.  Ulsan’s Administrative AI began with internal application, and its widespread impact on citizens remains unconfirmed.

When AI classifies subjects in tax audits, civil complaints, welfare, and safety, the impact of administrative judgments becomes greater than that of document support. Although the Personal Information Protection Commission released guidelines for Generative AI personal information in 2025 and Public AX privacy protection in 2026, Ulsan's procedures for AI impact assessment, model registration, and objection remain unconfirmed. As high-risk administrative AI expands over the next two to three years, errors will shift from simple response errors to disadvantages in taxation, support, and enforcement. While  the utilization of AI in Ulsan has begun, there is a lack of public evidence regarding the algorithm accountability framework.

When AI administrative assistants use external platforms, issues arise regarding the input scope of internal documents, civil complaints, and personal information, as well as whether training data can be reused. Although Ulsan has enhanced control over unauthorized individual use through official service contracts, data storage locations, access logs, and anonymization standards are not available in the publicly disclosed contract information. If usage increases by 2028 and control information is not disclosed, both efficiency and security risks will rise simultaneously.  The introduction of official platforms is the starting point for control, not proof of the completion of secure administration AX.

Ulsan City possesses a dedicated AI organization, a Smart City Division, a Big Data Center, a digital twin, and an AI administrative assistant. While the basic elements of organization, data, platform, and work tools have been established, the list of processes across all departments and the AX conversion rate are not disclosed. If readiness is measured by the number of systems over a period of two to three years, departments with no actual changes in work processes are included in the AX performance figures.  The foundational readiness for Ulsan City's administration AX is rated as 'High,' while the readiness for work redesign across all departments is rated as 'Unconfirmed.'

The budget for the AI ​​administrative assistant is 33 million won, and platform construction and subscription contracts are confirmed for 2026. While specific enough to start an official experiment, it differs significantly from the investment scale required to transform the entire administration of a metropolitan city, and the scope of use has not been disclosed. If the pilot budget level is maintained until 2028, it will be fixed as a productivity tool for a select group of users. The AI ​​administrative assistant is more of an initial demonstration project than a full-scale transition.

AI-related projects are being identified across various fields, including smart cities, spatial information, taxation, youth, and heatwaves. While the scope of application has expanded, common adoption standards across departments, reuse models, and comparative indicators for work time reduction are not being disclosed. If each department purchases AI as a separate project over the next two to three years, issues regarding redundant investment and data compatibility will intensify. Although  Ulsan’s administrative AI is expanding in terms of the number of cases, its common operating system remains incomplete.

While the establishment of a dedicated AI organization enhances central coordination, AX managers and performance indicators for each general department remain unconfirmed. In a structure where the dedicated department identifies projects and the operational units utilize them, the operational units' responsibility for problem definition and model improvement may be weakened. If departmental internalization does not occur by 2028, the volume of work by the dedicated organization will be disconnected from the overall level of AX within the administration.  Organizational diffusion is centralized, while operational internalization remains unconfirmed.

Ulsan carries out tasks related to public data opening, metadata management, data dams, and the joint utilization of big data and spatial information. Although the scope of data management is broad, the number of data combinations between departments, the correction period for quality errors, and the policy reuse rate are not disclosed. If operations are conducted primarily through openness and quality evaluations for two to three years, data not used for internal decision-making is also counted as performance. While  the expansion of data management has been confirmed, the expansion of policy reuse remains unconfirmed.

In the Ministry of the Interior and Safety's 2025 evaluation of public data provision, only 51% of all institutions received an "Excellent" rating or higher. While a gap remains between data openness and AI utilization at the national level, Ulsan's specific grades and improvement items are not clearly separated in the currently available data. Unless high-value AI training data is discovered by 2028, the performance of administrative AI will not improve solely through the opening of general statistics.  Ulsan's data administration has a quantitative foundation, but its AI training value still requires separate verification.

While Ulsan City Hall has introduced AI administrative assistants and digital twins, it remains unconfirmed whether the five districts and counties, as well as affiliated organizations, are using the same platform. Although citizen services move between metropolitan, local, and public institutions, the adoption of AI is highly likely to be fragmented by agency-specific budgets and systems. If only the main office undergoes advancement over the next two to three years, an AX gap will remain at the actual points of contact for civil complaints, welfare, and daily administration.  The AX implementation at the main office has commenced, but its expansion throughout Ulsan's administrative ecosystem remains unconfirmed.

While urban, industrial, and transportation issues transcend administrative district and departmental boundaries, data authority is divided by agency. Joint data and model operation performance data from Ulsan, its districts, and public enterprises are not disclosed. If agency-specific systems are maintained until 2028, citizens will repeatedly submit the same information, and AI will perceive this as incomplete data.  The gap limiting Ulsan's citizen-perceived AX is the data disconnect between agencies, rather than model performance.

The introduction of AI administrative assistants and the launch of a dedicated AI organization in 2026 marked the initial stage of Ulsan Administrative AX. While the performance and regulations of generative AI change every few months, local government budget, procurement, and security reviews operate on an annual basis. If the procurement cycle lags behind technological changes during the period of 2026–2028, the introduced systems will rapidly become obsolete during the contract period.  The time-sensitive risks for Ulsan Administrative AX stem not only from delays in implementation but also from the mismatch between fixed contracts and the AI ​​change cycle.

As AI city and industrial projects expand, tasks related to licensing, regulations, budgeting, civil complaints, and performance management also become more complex. If internal administrative AX lags behind industrial policy, organizations driving AI projects become bogged down in manual reporting and departmental consultations, yet changes in processing times are not disclosed. In two to three years, the speed gap between industrial AX and administrative AX manifests as investment delays.  Ulsan’s golden time for an AI city could be exhausted due to the speed of administrative decision-making rather than corporate support.

When administrative documents, queries, and revision histories accumulate on AI administrative assistants and external platforms, business processes adapt to the models and data structures of a specific provider. Subsequently, it is not disclosed whether prompts, business knowledge, and usage history can be transferred when switching platforms. If usage increases without standards for two to three years, it becomes difficult to switch providers despite issues regarding cost, security, and functionality.  Platform dependency in administrative AI is an irreversible risk that grows as usage accumulates.

When AI is used to classify taxation, welfare, enforcement, and civil complaints, past judgments become training data for the next model. If initial biases or erroneous administrative records are not corrected, they are repeatedly applied to the same groups and regions, yet no procedures for model correction or deletion are verified. If learning accumulates without an accountability system by 2028, errors will become fixed as administrative patterns rather than individual dispositions.  The irreversibility of administrative AI lies in the fact that erroneous judgments are reused as evidence for subsequent judgments.

Compression determination: Ulsan City Administrative AX is established only if reduced administrative steps, shortened processing time, policies changed or suspended based on data, reduction in citizens' repetitive documents, and algorithmic objection processing performance are confirmed before 2028, rather than the number of AI projects.

Ulsan City is assessed as being in the 'Administrative AI Introduction Stage' equipped with a dedicated AI organization, administrative secretary, tax AI, digital twin, and data platform , and the **'Administrative AX Pre-Stage' where inter-departmental work redesign, integrated citizen services, performance feedback, and algorithm responsibility are incomplete**.

Traces of the AI ​​City declaration are evident in industrial policies and organizational restructuring. However, there is a lack of integrated evidence demonstrating that Ulsan City's administration itself makes decisions more quickly with fewer steps, solves problems across departmental and agency boundaries, and returns the results to the citizens.

Currently, Ulsan's administration has moved toward an administration that uses AI, but it has not yet reached an administration where operational methods have been changed by AI.

Golden Time Judgment: AI Governance Initiation + Internal Administration Transformation Gap

Evidence Sources

• Ulsan Metropolitan City Administrative Organization Chart—AI Innovation Industry Office · AX Policy Officer
• Ulsan Metropolitan City Signs Contract to Build AI Administrative Assistant Platform
• Ulsan Metropolitan City, AI Administrative Assistant Service Subscription Agreement
• Ulsan Metropolitan City, AI Administrative Secretary Budget
• Ulsan Metropolitan City, AI-based Tax Source Discovery
• Ulsan Metropolitan City, AI Spatial Information Convergence Search Service
• Ulsan Metropolitan City Smart City Division—Data and Digital Twin Operations
• Ministry of the Interior and Safety, 2026 Joint Evaluation of Local Governments
• Personal Information Protection Commission, Public AX Privacy Protection Guide

Structural Insight — The success of a dedicated AI organization could delay AX across the entire administration

The dedicated AI organization concentrates budget, personnel, and technology to rapidly establish initial projects. Since general departments can request projects from the organization even if they do not directly handle AI, the number of cases increases immediately after the organizational restructuring. The structure is such that the performance of the dedicated organization is counted as the overall AI performance of the administration.

However, the departments best equipped with business knowledge are the operational divisions, such as tax, welfare, transportation, and environment. As dedicated organizations delegate problem definition and technology selection, operational departments remain merely users of AI models, failing to accumulate the capacity to correct data or redesign business processes when results are incorrect. This creates a paradox where the stronger the centralized AI capabilities become, the slower the internalization by operational departments becomes.

In this structure, even if a dedicated AI organization completes numerous projects, the overall administration maintains existing methods of approval, reporting, budgeting, and evaluation. AI accelerates the input and document production of outdated processes, and these faster outputs, in turn, generate more reviews and approvals. Automation may not reduce the volume of administration but rather expand it.

Therefore, the structural turning point for Ulsan City’s Administrative AX is not the size of the AI ​​organization or the number of projects. The actual completeness of Administrative AX is determined by whether it reaches a stage where each operational department directly manages its own problems, data, and results, even without a dedicated AI department.