K-water's AI Overhaul: Inside Korea's Smart Water Reform

Flood damage downstream of its dams in 2020 pushed Korea Water Resources Corporation — K-water — into adopting AI and digital twin technology in earnest, and by late 2025 it had set up a dedicated AI Division to pull the work together.

K-water works as a national water wholesaler: it manages dams, treats water at purification plants, and supplies municipalities. Its dam operations have long relied on technology, but forecasting leaned on historical climate patterns and decades of operator experience. As climate change delivers more frequent extremes of rainfall and drought, those methods increasingly fail to predict what's coming. AI became the answer to the forecasting gap.

What does the new division actually do?

The AI Division unites internal specialists from IT, administration, civil, mechanical, electrical and environmental engineering, plus telecommunications. Its job is to fold technologies built in separate silos — digital twins for water management, AI-powered treatment plants, smart water distribution network management — into a single framework spanning the full journey from dam to tap.

Ha-Dong Jung, VP of the AI Division, joined K-water in 1997 and became its first head. He frames the division as a natural continuation of the utility's history of technology-driven internal change.

That history mattered. K-water first collected dam levels, rainfall and water quality data for monitoring, then used it to forecast demand and plan production. Its ERP implementation refined metadata and data standards. None of it was done with AI in mind, yet those foundations let K-water start AI projects ahead of peers.

Why isn't the data AI-ready yet?

Accumulated data alone doesn't qualify. "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."

Connecting upstream rainfall to water quality, treatment output and supply requires joining data across facilities. K-water is investing in an ontology that defines data concepts and their relationships, combined with RAG (retrieval-augmented generation) so systems can find and use the right information.

Infrastructure also pushed back. Pulling large volumes of data for analysis sometimes crashed equipment that was never specified for the job.

"Sometimes we'd be told things were working, only to check and find that the process had stopped partway through," Jung says. "We're working internally to address those issues." The fix meant upgrading collection infrastructure and adding multiple data-acquisition routes. To consolidate scattered data, K-water has secured a site in Sejong for an AI data center it aims to build by 2030.

How do you give employees AI they'll actually use?

K-water built an internal service, K-water GPT, on Meta's Llama models, swapping in better models over time. But in-house development couldn't keep pace with the market.

"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."

Because the internal work network and internet are separated, employees initially couldn't use internal documents with AI; K-water is building connections to make external generative AI reachable from the work network. Broad early access to encourage adoption later ran into cost, so the company now sets individual usage limits.

What still isn't solved?

The hardest gap: pace. A typical IT project takes six months to a year to prepare, about a year to implement, and six months to close out — while AI models keep changing throughout.

"Things have changed by the time you finish preparing, and they change again while you're working your way through implementation," Jung 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 internal capability is slow too: hiring staff and completing project reviews and budget assessments takes considerable time, a continuing structural challenge for the public sector.

Still, K-water is regarded as one of South Korea's leading public enterprises in AI, and OpenAI has highlighted it as a prime example of AI use in disaster response. Jung credits persistence: earlier efforts build on one another, and changing technology and organizational thinking can make previously impossible initiatives feasible.

The division's goal is establishing K-water as a global water company leading climate-change response — and Jung says the AI work so far is only the beginning, with his role being to remove obstacles when employees hit walls. As AI project cycles keep shrinking against multi-year infrastructure timelines, K-water's experiment in-house capability may become the template utilities everywhere end up studying.