Korea Water Resources Corporation, better known as K-water, serves as sort of a water wholesaler, managing water in dams, treating it at purification plants, and supplying it to local municipalities. Now, the company is using AI to transform how to connect the flow of water from one end to the other.
K-water has a long history of using technology to operate dams. In the past, however, it relied largely on historical climate patterns and operating experience accumulated over many years. As climate change brings more frequent and unexpected extremes of rainfall and drought, more situations are proving difficult to predict using those methods. So K-water needed new tools for analysis and forecasting, and AI was the answer.
Source of the matter
Flood damage downstream of dams in 2020 prompted K-water to begin adopting AI and digital twin technology in earnest. To extend that work across the organization, it established the AI Division in late 2025, which brings together internal specialists from a range of disciplines, including IT, administration, civil engineering, mechanical engineering, electrical engineering, environmental engineering, and telecommunications.
Its task is to bring technologies developed in separate areas — including digital twins for water management, AI-powered treatment plants, and smart water distribution network management — into a single framework. That framework spans the entire process of managing water in dams, treating it, and delivering it to users.
Ha-Dong Jung, VP of K-water’s AI division, joined in 1997 and became its first head. He’s witnessed its IT evolution firsthand, and to him, the division’s creation was a natural progression. K-water has consistently used technology to drive internal innovation, and AI is a continuation of that work.
Data management has been a particular focus. Initially, K-water collected information on dam levels, rainfall and water quality mainly for monitoring. It later began using that data to forecast water demand and plan production and supply. Through initiatives such as its ERP implementation, it also refined its metadata and data standards. Those efforts weren’t undertaken with AI in mind, but the foundations they established allowed K-water to begin AI projects ahead of others, and now they underpin the AI Division’s work. Collecting large amounts of data, however, didn’t make it immediately ready for AI.
“In the past, the purpose of our data work was to collect and store data and display it so people could understand it at a glance,” Jung says. “Today, that accumulated data is just a collection of numbers. To use it for AI, we need to define what each piece of data means and how the different data points relate to one another.”
Understanding how rain upstream of a dam affects water quality and output at a treatment plant, and ultimately the water supply, requires connecting data from multiple facilities. K-water is therefore investing in an ontology that defines data concepts and their relationships. It’s also combining this with RAG to build a system that can find and use the information needed.
Not always calm waters
There’ve been practical obstacles, says Jung. Attempts to extract large volumes of data for analysis sometimes caused existing equipment to stop working. K-water had to improve its collection infrastructure, including adding multiple routes for acquiring data.
“The existing equipment was designed with specifications suited to its original tasks,” he adds. “It wasn’t built to accommodate the extraction of the large volumes of data that AI projects require. Sometimes we’d be told things were working, only to check and find that the process had stopped partway through. We’re working internally to address those issues.”
To bring scattered data together and use it more efficiently, K-water has secured a site in Sejong for an AI data center, which it aims to build by 2030.
Bringing workplace AI up to employee standards
Alongside its facility operations, K-water is changing the AI tools available to employees. It has actively explored ways to provide external services alongside AI services developed in-house. As a public enterprise, it faces constraints on adopting external services quickly, but it’s worked through those restrictions step by step.
K-water first built an internal service called K-water GPT using Meta’s Llama models, improving performance by replacing the underlying models over time. But there were limits to how closely an in-house service could keep pace with rapidly advancing external AI offerings.
“Employees already judge AI against global services such as ChatGPT, Gemini, and Claude,” Jung says. “It’s difficult to keep up with the pace of development in global AI models through in-house development alone, so we introduced external services that allow employees to choose among major AI models as needed in an environment with enhanced security.”
Further improvements followed. Because its internal work network and internet network are separate, employees faced limitations in using internal documents with AI. K-water began working on connections that would make external generative AI services available from the internal work network as well.
Growing adoption also brought the challenge of managing costs. Initially, K-water allowed broad use to encourage employees to try the tools. It now sets individual usage limits.
One unresolved challenge is the gap between the pace of technological change and the time it takes to execute projects. According to Jung, an IT project typically takes six months to a year to prepare, about a year to implement, and another six months or so to close out. AI models and capabilities continue to change throughout that period as well.
“Things have changed by the time you finish preparing, and they change again while you’re working your way through implementation,” he says. “Over the longer term, I think we may need to create an environment where our own employees can respond directly, rather than continuing to rely entirely on outside contractors.”
Building that internal capability is no simple matter either. Securing staff and completing project reviews and budget assessments takes considerable time. Meanwhile, technology moves on, and how to adapt procedures to the pace of technology is a continuing challenge for the public sector.
Past efforts paving the way for progress
Public enterprises and agencies are often seen as slow to respond to technological change because of the procedures they must follow. Structural constraints do make it difficult to accelerate change. Yet K-water is regarded as one of South Korea’s leading public enterprises in AI. OpenAI has also highlighted it as a prime example of AI use in disaster response, drawing attention to its water management capabilities.
Jung attributes that progress to a willingness to embrace change and take on challenges that have become ingrained in K-water’s employees.
“Earlier in my career, I found it difficult to deal with frequent system changes,” he says. “If a project of high cost fell short of expectations, I considered it a failure. But over time, I saw the people who came after me build on the technology and experience from those projects, try again, and eventually deliver results. That’s when I realized that tech doesn’t end with a single attempt. Earlier efforts build on one another. I think K-water employees have that DNA of persisting and pushing ahead.”
That doesn’t mean repeating the same work. Advances in technology, shifts in organizational thinking, and changes in available resources can make previously difficult initiatives possible. K-water’s AI projects, too, were made possible by the operational data it had accumulated over many years.
Jung says the progress so far is only the beginning, with the company’s use of AI in the early stages.
A measure of success
The AI Division’s ultimate goal is to help establish K-water as a global water company leading the response to climate change. In pursuing that goal, Jung sees his role as helping employees continue to take on challenges. When they hit a wall, he examines the problem with them and looks first for obstacles he can remove as a leader.
“I need to start by asking whether there’s something I can do to help,” he says. “If it’s a problem I can resolve at my level, I do everything I can to resolve it. If it isn’t, I spend more time listening to what employees have to say.”