South Korea’s construction industry is not adopting artificial intelligence as quickly as expected, largely because jobsite data are scattered in nonstandard formats and rules remain unclear, industry officials said. Analysts also point to a lack of supporting infrastructure as another obstacle to AI-driven change.
Industry sources said on the 22nd that many domestic construction sites still lack the data foundation AI systems need to learn and make decisions. Design documents, photos, inspection logs and process data are often stored in different formats, making them difficult for systems to read and use at once.
Companies sometimes describe such unstructured material as data “piled up like a stack of drawings.” To improve training data quality, they say, the first step is standardizing dispersed, unstructured records — a task that has fueled calls for government-level guidelines. The Korea Research Institute for Construction Policy said AI bottlenecks include difficulties in standardization and data accumulation, along with fragmented data management. It recommended pursuing a public-private data linkage platform, improving data consistency, establishing security systems and advancing institutional reforms.
While construction and engineering firms are trying to expand AI use, efforts often remain at the individual level rather than becoming organizationwide change. According to the Korea Technology and Information Promotion Agency for SMEs, 80% of construction companies with annual revenue of 500 million won or more do not even collect data, underscoring the sector’s low data maturity.
Regulation has also struggled to keep pace with changes on the ground. South Korea’s “Framework Act on the Development of Artificial Intelligence and the Establishment of a Foundation of Trust” is in effect, but an AI implementation framework tailored to construction remains unclear. Issues repeatedly raised at construction sites — including data responsibility, liability when AI makes an incorrect judgment, and standards for pilot applications — still require more detailed follow-up rules.
Jeon Young-jun, a research center director at the institute, said much of the data held by construction companies is stored as scanned PDFs of drawings or documents, which is not suitable for AI training. “It looks like text on the surface, but it is actually an image, so AI has difficulty recognizing and learning it properly,” he said.
He said data must be structured and standardized first, and called for government standards for building a common data environment, or CDE. He added that companies should develop technology under those shared standards.
Overseas, results are already emerging in administration and site management. Qatar officially announced last year that it introduced an AI-based intelligent building permit system, cutting average permit processing time from 30 days to about 120 minutes. The system automatically reads design documents and checks compliance, reducing review backlogs. The case has drawn attention because lower permitting uncertainty can make start dates easier to forecast and speed up financing such as project finance, or PF.
Singapore is also accelerating jobsite digitalization. The Building and Construction Authority, or BCA, has identified data-driven collaboration and digital transformation as core tasks in its industry transformation strategy. Its Site Management Platform, or SMP, serves as a central digital hub that gathers and manages information generated at sites. BCA is linking SMP with its Site Management Data Standard, or SMDS, to standardize safety, productivity, quality, schedule and cost data — an effort aimed at aggregating site-generated databases into a common data environment and laying the groundwork for AI use.
At home, industry participants cited expanding regulatory sandboxes and preparing in advance for predictable risks as key steps to improving AI infrastructure in construction.
An industry official said the demonstration process should be simplified so technology can take hold at sites first. The official said adoption will continue to lag unless standards are set in advance for AI misjudgments and safety issues, protection of personal information and trade secrets, and responsibility in procurement and permitting processes.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.