We recently shared the big-picture policy questions around AI and water. Now here is the practical follow-up: what does AI in the water sector actually look like on the ground? The Water Center at the University of Pennsylvania has cataloged real, working examples, and the answer is more grounded, and more useful, than the hype suggests.
Their new report, “AI for Water: Examples in Action,” released in September 2026, is a plain-language tour of how water utilities are already putting AI to work. It sorts the examples into four purposes: saving time, improving operations, improving planning, and improving public decision making. A few highlights stood out to us.
AI that saves time
The easiest wins are administrative. A Pennsylvania water utility serving 116,000 customers now uses AI to read and route incoming emails to the right department. Others use internal chatbots so staff can pull answers from procedures and training documents in seconds, and tools that draft permits and review applications, cutting the busywork so people can focus on judgment calls.
AI that improves operations
This is where AI has the longest track record. Utilities use predictive platforms to forecast demand, equipment wear, and compliance risk before problems happen. Digital twins, detailed computer models of real systems, are helping in striking ways: Seattle City Light built one of its Diablo Dam to plan maintenance, and Singapore’s national water agency used an AI-powered “anomaly leak finder” that caught two real leaks in its first three months. On the treatment side, machine learning is fine-tuning chemical dosing well enough to cut chemical use by 10 to 30 percent while staying in compliance.
AI that improves planning
From smart meters that flag leaks to acoustic sensors listening to aging pipes, AI is helping utilities see their systems more clearly. It is also reshaping big projects. Sacramento’s regional sanitation district used a digital twin to coordinate a 22-project plant upgrade and brought the cost down from $2.1 billion to $1.7 billion, money it redirected into a recycled water program for farms and habitat.
AI that improves public decision making
This is the category closest to our mission. The report highlights WaterMark, a free, citation-backed tool that lets anyone, residents, regulators, or local governments, estimate the water impact of a proposed data center: daily consumption, watershed stress, and more. For a network that has spent this year tracking data centers reaching for our rivers, a public tool that quantifies exactly what a project would take is the kind of transparency that changes conversations. Other tools track water policy decisions and help rural communities lead their own land and water planning, guided by a philosophy one project sums up as “more participation, less AI.”
On the horizon
Among the emerging tools, one caught our eye: researchers at Drexel University built a system that detects lead service lines from the surface using acoustic waves and deep learning, with an 83 percent success rate in blind testing and no need to dig up a homeowner’s yard. With Philadelphia alone facing 20,000 to 25,000 lead lines to replace, that is a technology worth watching.
The report’s honest bottom line matches our own view. AI is already delivering real value in water, but small and medium utilities risk being left behind without access to the data, training, and funding to adopt it. Making sure the tools reach the communities that need them most is the whole game.
A note from the H2O Water Network
The H2O Water Network uses generative AI that is off the grid, private, and green. If your organization wants the benefits of these tools without sending your data to the cloud or adding to the water and energy footprint that comes with it, ask us how.
Ask Us HowSummarized from “AI for Water: Examples in Action,” Version 1.0 (The Water Center at the University of Pennsylvania, September 2026). Examples are drawn from the report and are illustrative, not endorsements.
