In This Series

Part 1 — Data as National Infrastructure
Part 2 — How Data Becomes an Industrial Asset: Yang Jae-soo’s Framework for Verification
Part 3 — DataStreams Takes Trusted, AI-Ready Data Ecosystems Global
Part 4 — FLock.io CEO Jiahao Sun on Keeping Data Local While AI Learns
Part 5 — Causal Dynamics Lab CEO Hasibul Haque: “AI Agents Need Operational Context and Accountability”
Part 6 — KIT Analysis: Korea Must Prove Its Data Competitiveness Through Verifiable Industrial Outcomes

AI Outcomes Depend on Data Flow and Control

For AI models to deliver results in industry, businesses and public institutions need to connect data scattered across multiple operational systems into a coherent, trusted flow. If the meaning, quality and usage rights of that data remain unclear, organizations may struggle to explain an AI system’s reasoning or determine the cause and responsibility when errors occur. An operating framework that tracks where data originated, how it changed and who used it is the foundation for turning AI into a verifiable service.

DataStreams is positioning data fabric as a foundation for enterprise and public-sector AI, linking data integration and standardization with quality management, metadata, lineage and access control. Lee Young-sang, CEO of DataStreams, said a “trusted data infrastructure layer” connecting AI models with organizational data will shape the AI competitiveness of companies and countries.

The company recently won a $2.6 million contract for the Pakistan Post Automation Project. It plans to apply data management technologies developed in Korea to Pakistan’s public infrastructure while transferring technical expertise, operating methodologies and workforce-development capabilities. The approach follows what DataStreams calls “Core Standard + Local Adaptation,” under which the common data infrastructure is standardized while laws, policies and operational procedures are adapted to local conditions.

This interview examines why public institutions need to establish data governance before deploying AI and how data sovereignty can be converted into domestic technological capability. It also considers what would be required to extend the Pakistan project to other markets and turn DataStreams’ experience in Korea into verifiable results overseas.

The following is a Q&A with DataStreams CEO Lee Young-sang.

Q1. Many countries are investing in AI models, but public-sector data often remains fragmented or incompatible across systems. Why should emerging economies establish data integration, quality management and governance before deploying AI at scale?

As competition in AI intensifies, many countries are seeking to invest first in AI models or GPU infrastructure. Yet the performance of AI ultimately depends on the reliability of the data from which it learns and to which it refers.

Within the public sector, information about the same citizen, company or region often exists across multiple agencies and systems in different formats and according to different standards. The same data field may be described differently from one system to another, while its origin may be unclear or its quality unverified.

AI can technically generate an answer under these conditions. Whether governments can trust that answer enough to use it in policymaking or public services is an entirely different matter. Public institutions need to be able to trace the data behind an AI decision and identify the cause when the system produces an incorrect result.

Before adopting AI, institutions therefore need to establish foundational data governance covering integration, standardization, quality management, metadata, data lineage and access control. I compare this to constructing a building. If the AI model is the building, data infrastructure is the ground and foundation that support it. Increasing the height of a building on weak foundations is not sustainable.

In the public sectors of emerging economies, awareness of and investment in data have often developed slowly and at a limited scale. As a result, much of the available data has not been prepared in a form that AI can use. Designing the data foundation at the beginning of a project can keep the initial investment relatively modest. Correcting the structure after systems have become complex and entrenched can be so costly that the project itself becomes difficult to pursue.

Emerging economies should prioritize policies that turn their national data into a trusted national asset before rushing to deploy large-scale AI models. In the AI era, data is becoming critical national infrastructure, comparable to financial, telecommunications and electricity networks in its importance to national competitiveness.

Q2. Public-sector data is closely tied to national security, privacy and data sovereignty. What cooperation model can DataStreams offer governments seeking to retain control over their data while building domestic technical capabilities?

Data sovereignty involves more than storing data on servers located within a country. National control over data can be considered in terms of both production and utilization, with sovereignty becoming particularly important at the utilization stage. Controlling data rights, storage environments and access provides sovereignty at the production stage. Genuine data sovereignty also requires technologies that a country can control to manage data movement, verification, integration, visualization, analysis and forecasting.

A data center can be built with technologies from the United States, China or other countries, and strong security controls can help ensure that the data remains within national borders. Data must still pass through a “data utilization layer” before it can be put to practical use. The data fabric technologies required for this layer often depend on products supplied by U.S. companies or on open-source software.

Even when local professionals learn to operate these technologies, a country may still confront sovereignty issues in the use of data and AI if it cannot internalize and further develop the underlying capabilities. A model in which an overseas supplier monopolizes all technology and operations is therefore undesirable for public-sector projects over the long term.

DataStreams seeks to help local governments and institutions acquire technical knowledge and operational capabilities alongside the technology itself. Data should, wherever possible, be managed within its country of origin in accordance with local regulations and policies. Systems should also support environments including on-premises infrastructure and private clouds when required by the government.

It is equally important to design systems around open standards and interoperability so that institutions do not become dependent on a single global platform. Drawing on 25 years of data management experience, DataStreams plans to share data-management and utilization technologies with partner countries while supporting education and workforce development. This could help data fabric capabilities develop into a domestic industry.

The process can connect digital transformation and AI transformation by combining data technology, AI technology and AI workforce development. What needs to be exported extends beyond software licenses. Countries need a capability-transfer model that brings together technology, operating methodologies, data management frameworks and talent development. This offers a practical basis for building long-term trust with countries that regard data sovereignty as integral to digital transformation.

Q3. Public-sector digital transformation cannot be completed through technology exports alone. What kinds of partnerships and workforce-development systems are needed to build sustainable data ecosystems involving public institutions, companies, universities and government agencies?

One major reason public digital-transformation programs fail is that technical knowledge disappears when a project ends. Sustainable digital transformation should therefore extend beyond system deployment and create a local technology ecosystem.

Governments and public institutions need to establish data policies, standards, security requirements and governance frameworks. They should also expand digital- and AI-transformation projects that help local technology companies develop implementation and operating capabilities. Universities and training institutions need to supply professionals required by industry, including data engineers, data architects and specialists in data quality and governance.

The role of global technology companies also needs to evolve. Instead of performing every function themselves, they should initially provide core technologies and methodologies, then enable local partners to use those capabilities to develop new services.

One possible approach is for local and Korean engineers to build systems together and subsequently establish local training programs through a train-the-trainer model. Universities could develop courses in data management and AI utilization, while certification systems could recognize companies and engineers that reach defined levels of capability.

Such efforts could draw on national budgets and international cooperation funds, supplemented by joint projects or joint ventures designed to help local technology companies grow. When domestic companies participate in the value created by a project, more of that value remains within the country and investment in education can translate into an industrial ecosystem.

Under this model, local companies and professionals can continue delivering new projects after the initial public-sector program has ended. Embedding technical capabilities locally allows the ecosystem to keep developing and make a substantive contribution to national digital transformation.

Q4. DataStreams recently won a $2.6 million contract for the Pakistan Post Automation Project. How does the project demonstrate the potential for Korean data technology to expand into overseas public infrastructure, and what does the “Core Standard + Local Adaptation” model involve?

The significance of this project extends beyond the export of Korean software. It marks the use of data management technologies developed in Korea to modernize another country’s core public infrastructure.

The Pakistan Post Automation Project will digitize postal services nationwide and establish a foundation for connecting and using data generated across logistics, finance and other public services. In a large public system, reliably integrating diverse legacy systems and data — while maintaining quality and consistency — is as important as creating new services.

Pakistan has developed substantial capacity in IT services and outsourcing, although its data market remains relatively underdeveloped. DataStreams plans to implement the postal modernization project on a data-driven basis, work with local partners to expand the data market and build the foundation for more effective AI services.

Over the past 25 years, DataStreams has developed expertise in data integration, quality, metadata and governance across Korea’s financial, public and manufacturing sectors. The Pakistan project is an important case for demonstrating whether this experience can be applied to public infrastructure beyond the Korean market.

The company does not intend to treat the project as a one-off deployment. Its goal is to transfer technical and operating capabilities so that local institutions and businesses can operate and further develop the system themselves. DataStreams then aims to extend cooperation into other areas of public service, including finance, logistics and digital government.

Pakistan’s strong English-language documentation capabilities and close relationships with the Middle East could also make the country a bridge to neighboring markets. For DataStreams, Pakistan represents a starting point from which to demonstrate that Korean data technology can support public digital infrastructure internationally.

Across Asia, the Middle East and Africa, post offices serve as important points of contact between citizens and financial, logistics and government services, as well as delivering mail. The digital transformation of postal systems can therefore connect directly with the wider modernization of public services.

Some components can be standardized across countries. These include data-integration architecture, quality management, metadata management, standardization, lineage, security, access control and real-time data connectivity. Converting these capabilities into a common framework would allow institutions to deploy proven data infrastructure while reducing implementation time and risk.

Other elements require localization, including privacy laws, financial regulations, data-retention policies, postal and logistics procedures, language, government structures and legacy systems. DataStreams describes its global model as “Core Standard + Local Adaptation.” The core data infrastructure is standardized, while policies and operating processes are adapted to each country.

If the model is sufficiently validated in Pakistan, it could be extended to postal and logistics systems in Southeast Asia, the Middle East and Africa, as well as to digital-transformation projects in finance, public administration and social services.

Q5. If global AI competition is shifting from model scale toward the ability to use trusted data, what conditions should data infrastructure meet, and what strategic position could Korea and DataStreams occupy?

I expect the focus of AI competition to shift from who owns the largest AI model to who can provide the most trusted data in a form that AI can use.

Most governments and companies already hold vast amounts of data. The difficulty is that the data remains distributed across different systems, with inconsistent meaning and quality, and is not managed in a form that AI can use safely.

Korea could establish an important position in this field. It has accumulated extensive experience in systems integration and data management by operating large and complex IT environments across finance, digital government, telecommunications and manufacturing. This experience could provide a practical model for countries that need to transition gradually toward AI while keeping existing systems in operation.

DataStreams is not seeking to compete with global technology companies on the scale of AI models. It is focusing on the trusted data infrastructure layer that connects diverse data sources with AI. Data needs to be visible, standardized and subject to quality controls, while its meaning and relationships must be consistently understood.

Organizations should also be able to trace who used which data, under what authority and for what purpose. The infrastructure needs to work with multiple technologies without locking users into a particular cloud provider or AI model.

The core of DataStreams’ work has remained largely unchanged over the past 25 years. It defines the technical and operational meaning of data, improves standards and quality, integrates information from different systems and manages its meaning and flow so that organizations can use it with confidence. These capabilities once provided the foundation for data warehouses and big-data systems. They now underpin AI services.

For DataStreams, the AI era is therefore less an entirely new market than a period in which the relevance of its accumulated data integration and management technologies has become more visible. The company’s global strategy currently follows three directions.

First, it is connecting data standards, quality, metadata, governance and lineage into a common foundation that turns corporate and government data into AI-ready data that systems can use with confidence.

Second, DataStreams is developing these technologies into an intelligent data fabric platform that links diverse repositories, cloud services and on-premises systems through a virtualized integration layer. The platform is intended to work with both general-purpose and domain-specific AI technologies. The company says this approach can preserve freedom of choice, reduce dependence on a single global platform and address some limitations of open-source platform architectures through its proprietary technologies.

Third, DataStreams is seeking to build a global partnership model through overseas projects, including the Pakistan initiative. This involves joint implementation and operation with local companies, technology sharing and collaborative workforce development. The company is also focusing on data capabilities in defense and manufacturing, where it sees significant national implications.

DataStreams summarizes the requirements for intelligent data fabric infrastructure in three terms: trusted, sovereign and AI-ready. Data must be reliable, remain under the control of the country or institution that owns it, and be prepared for immediate use by different AI systems. These principles are intended to define the technical conditions for global data infrastructure rather than merely describe individual DataStreams products.

Korea has an opportunity to position itself as a data infrastructure partner that helps AI systems around the world use trusted data. DataStreams aims to draw on Korea’s data management experience to establish an infrastructure model that allows countries to retain control over their data while realizing its value through AI.

Q6. What are the principal domestic and overseas deployments of DataStreams technology, and what verifiable results have they produced?

DataStreams’ data fabric comprises 14 component technologies, each of which has been used in enterprise-level deployments.

The company has built data integration, standardization, quality management, metadata, master-data management, data catalogs and governance frameworks for organizations across the financial, public and service sectors. Recent customers include the Korea Racing Authority, Hana Life Insurance, KB Securities and CESCO.

At the Korea Racing Authority, DataStreams deployed its DeltaStream change-data-capture system and TeraONE extract, transform and load solution to establish data-standardization guidelines. The company said it applied phased validation while cleansing and migrating information from multiple source systems to improve data quality.

It also built a sales data warehouse and summary data marts to support operational use of data. An integrated customer data platform provided the foundation for a single view of customer journeys and histories.

At Hana Life Insurance, DataStreams deployed IRUDA, MetaStream and QualityStream to support data governance and the standardization of information systems for the insurer’s AI and next-generation system development. The project established a data-governance framework, reorganized information systems around common standards and created the basis for end-to-end data-flow management, self-service business intelligence and management dashboards.

At KB Securities, the company used MetaStream and QualityStream to improve data standardization and quality management. It automated recurring work such as standard-field inspections, enhanced personalization and the user interface and experience, and sought to improve data quality and consistency across development, validation and production environments. The project also reorganized data structures and operating procedures to establish an AI-ready data foundation.

At CESCO, DataStreams deployed IRUDA, MetaStream, MetaStream for Bizdata and Q-TRACK to establish data governance, IT and business metadata management, data-flow management and a data catalog. The project established a framework covering the data lifecycle from creation and use through disposal. It also made business terminology, standard indicators, data flows and quality information searchable and usable on an end-to-end basis. Quantitative improvements in the CESCO project can only be assessed after the systems have been in operation, according to the company.

DataStreams says these projects show how its technology can connect integration, standards, quality, data flows and metadata to improve operational usability and readiness for AI. Independent, comparable performance indicators will nevertheless be needed to assess the scale of those improvements across deployments.

Turning Domestic Experience into Global Data Infrastructure

Lee’s answers indicate that DataStreams’ international strategy extends beyond selling individual software products. Its proposed model combines data standards and quality controls with operating methodologies and workforce development, helping local governments and companies use AI while retaining control over their data.

The Pakistan Post Automation Project will be an early test of this strategy in overseas public infrastructure. The next challenge is to present objective evidence from domestic and international deployments, including improvements in data quality, reductions in preparation time, user adoption, operating costs and measurable AI-service outcomes. Korean data technologies will be better positioned to compete in international standards and public-procurement markets when they can demonstrate verifiable operating results.

The next installment features Jiahao Sun, founder and CEO of FLock.io, and examines the potential and limitations of federated learning, which allows multiple institutions to train AI models collaboratively without transferring raw data outside their own environments. It will consider how organizations can pursue joint AI development while maintaining control over sensitive data and protecting data sovereignty.

Korean version available here

Monica Younsoo Chung More by this reporter