Shin Jang-ho, head of the Korea IT Service Industry Association, laid out policy directions to improve the structure of public software (SW) projects and pivot for the age of artificial intelligence (AI). Because the IT service industry is currently stuck in a revenue structure where it is hard to be paid commensurate with the work performed, the plan is to improve the ordering system and establish new compensation standards suited to the AI era.
At a press briefing held Apr. 29 at El Tower in Seocho-gu, Seoul, Shin said, "We must establish a compensation system tailored to the AI era and implement a flexible contracting regime for public SW projects." Shin, who serves as CEO of ITCEN ENTEC, was appointed as the new head of the Korea IT Service Industry Association in Feb.
In the software sector, companies have long struggled with the problem of not receiving additional compensation even when tasks increased during the bidding and execution of public projects. Although the system requires a task review committee to deliberate changes, in the field the committee either does not convene or meets only formally in many cases.
To resolve the issue, lawmaker Lee Hae-min of the Rebuilding Korea Party recently proposed an amendment to the Software Industry Promotion Act. The bill mandates convening the task review committee in the absence of special reasons and requires securing the resources needed for follow-up measures based on the committee's findings.
While welcoming the amendment, Shin also said the request for proposal (RFP) should clearly reflect calculation standards such as function points (FP). Currently, because project scope and calculation standards are not presented clearly, it has been difficult to assess increased work objectively during execution.
He also stressed the need to establish a legal basis to allocate additional resources when tasks change. Pointing to the limits of the current aggregates lump-sum bidding system, Shin raised the need to shift to itemized bidding. Under lump-sum bidding, even if tasks increase or expense rises during execution, the amount cannot exceed the figure set at contract signing. Even if the task review committee recognizes the need for changes, in practice it is difficult to increase the actual project budget.
By contrast, itemized bidding calculates expense by detailed line items, allowing cost adjustments when designs change or tasks are added. It is a method mainly used in construction, where changes in labor and raw material expense are reflected.
"In large public projects, deficits can reach as high as 50% of the budget," Shin said. "Excessively expanding tasks must be paid fairly."
The need to make compensation standards for AI projects more realistic was also raised. Currently, SW development projects calculate compensation by function points, but it is hard to apply this to AI projects such as fine-tuning large language models (LLM) and building retrieval-augmented generation (RAG). Because continuous updates and operations are essential, critics note that it is difficult to calculate appropriate compensation under the existing approach.
In response, the association is pursuing research to establish new compensation standards applicable to AI projects.
He argued that, in the long term, the revenue structure of the IT service industry must be improved. According to the association, as of 2024, the average profit margin of IT service corporations is 8.1% for large corporations, 5.3% for mid-sized corporations, and 1.7% for small corporations. In particular, the majority of small and mid-sized companies rely on public projects, which account for 30% of the domestic market.
"Public-sector IT budgets remain in the 1% range domestically, compared with about 3% in major countries," Shin said. "Budget increases are needed to strengthen industrial competitiveness."