South Korea is moving to turn public procurement into a launchpad for artificial intelligence startups, in a bid to ease one of the sector’s most persistent barriers: the need for previous government sales before winning a first government contract.
The shift matters because the state is already a major buyer. According to the Public Procurement Service, public contracts reached KRW 238.5 trillion in 2025, equal to about 8.9% of nominal GDP. That scale gives policymakers a ra...
Continue Reading This Article
Enjoy this article as well as all of our content, including reports, news, tips and more.
By registering or signing into your SRM Today account, you agree to SRM Today's Terms of Use and consent to the processing of your personal information as described in our Privacy Policy.
re chance to use purchasing power not merely to buy equipment and software, but to help promising AI firms prove themselves in live public-sector settings.
Wooju Jung, chief executive of inDJ and a member of the Presidential Committee on Artificial Intelligence Strategy, argues that grants alone are not enough. In a written interview with KoreaTechDesk, he said governments should act as early customers rather than only as funders.
“I believe the most effective way for governments to support AI startups is not simply by providing more grants, but by becoming their first customer.”
He added that procurement itself needs to be redesigned for the AI era, rather than merely expanded.
The policy response in Seoul has been to soften long-standing entry barriers. Under previous rules for AI products on Korea’s Multiple Award Schedule, suppliers generally had to show at least KRW 30 million in delivery performance, meet a minimum supplier threshold of three, and follow standard specifications. The Public Procurement Service has now removed the delivery-performance requirement for eligible AI products, lowered the supplier threshold to two, allowed supplier-proposed specifications, and waived the contract-stage eligibility review. Certain public purchasing qualification reviews can also award an extra 1.5 points to AI products.
The changes are designed to break what startups describe as a circular problem: the need for public-sector experience before public-sector access. Jung said many strong young companies are kept out simply because they lack a long sales record or prior government contracts.
There is still a balancing act. Public buyers need assurance on continuity, security and performance, especially when public money and essential services are involved. But Jung argues that technical evidence should count for more than institutional history. He has pointed to patents, venture certification, national R&D success, recognised AI benchmarks and independent technical assessments as better indicators of capability than procurement history alone.
Seoul has also made it easier to buy AI products that have already entered the system. The threshold that triggers second-stage competition rose sharply, from KRW 50 million to KRW 200 million for general products and from KRW 100 million to KRW 400 million for SME-competitive products. That means public bodies can move more quickly to purchase qualifying AI products below those levels without another layer of competition.
For Jung, that speed is not a procedural luxury but a precondition for innovation. He has argued that governments should make wider use of simplified or direct contracts for pilot projects within appropriate budget limits, saying faster procurement allows faster validation and better services.
The reform also comes alongside a more technology-specific assessment model. The Ministry of Economy and Finance has said an AI-focused Innovation Product review track is expected around September 2026, beginning with the fourth supplier-proposed designation round. Under that framework, public-value checks are expected to focus on data security, personal information protection and system stability, while the AI assessment itself will centre on technical excellence, risk management, reliability and suitability of the model.
That approach reflects a broader change in thinking: if procurement is meant to reward innovation, the evaluation process has to judge innovation on its own terms. As Jung put it, AI startups need not only money, but places where they can test and improve their systems.
He said cities, agencies and public institutions can serve as “living laboratories” for new technology. Korea’s 2026 Government First Demonstration and Purchase Project appears to follow that logic, linking public problems with startup solutions in real-world settings. In the smart-city round, 32 challenges were proposed by 18 institutions and 28 demonstration projects were selected after assessments of feasibility, collaboration and scalability. Among them were a conversational sign-language AI system for bus stops, proposed by Korea Land & Housing Corporation, and a vision-language-model CCTV tool for detecting anomalies in solar facilities, proposed by Incheon Port Authority.
The state is putting more money behind the idea as well. The Public Procurement Service has increased its 2026 Innovation Product pilot-purchase budget to KRW 83.9 billion from KRW 52.9 billion in 2025. But the harder question is whether that spending reaches firms that previously could not get through the door.
Jung believes it should. He says government should reserve a meaningful share of procurement for innovative startups and widen opportunities for regional firms to run practical pilot projects. The goal, in his view, is not simply more public spending on innovation, but a market in which early-stage companies can win real buyers.
That logic is not unique to Korea. The European Commission has long described innovation procurement as a way for public buyers to open markets, widen access for smaller firms and act as early adopters of technologies not yet widely available. In that sense, a public contract does more than generate revenue: it provides validation, reference value and credibility.
South Korea’s reforms suggest that policymakers are beginning to see procurement in that light. The real test will be whether technically credible AI startups with little or no government sales history can now win public contracts on the strength of their technology rather than their past.
Source: Noah Wire Services
