.png)
If you are a utility leader or GIS manager working toward the Lead and Copper Rule Improvements (LCRI) deadlines, your desk is likely piled high with data. Between historical tap cards, water main logs, and the looming November 1, 2027, baseline inventory deadline, the sheer volume of "unknowns" can feel overwhelming.
(For a deeper look at why reducing unknowns is critical and how software and modeling can support that effort, explore the following articles: Preparing for Lead and Copper Rule Compliance: Reducing Unknowns, Your LCRI Journey to Clean Water.)
We’ve always believed that digging up every single service line in a community isn't just incredibly disruptive, it’s also an inefficient use of ratepayer dollars when modern data science may assist with the heavy lifting.
That’s why a recently released draft fact sheet from the EPA ("Service Line Inventory Tips") caught our eye. If you've been sitting on the fence about whether federal regulators will embrace data driven methods to solve the unknown service line challenge, the EPA just sent a clear and favorable signal.
Let’s break down what the EPA said, what it means for your utility, and how to navigate the important role your state primacy agency plays in the approval process.
The Federal Green Light for Predictive Modeling
For those of us who spend our days building predictive models for water infrastructure, reading the EPA’s latest draft guidance was incredibly refreshing. The agency doesn't just passively permit predictive modeling; they actively champion it.
The fact sheet explicitly states that the EPA "encourages water systems to use, and primacy agencies to allow the use of predictive modeling techniques." Even more exciting for the future of our industry, the EPA explicitly validated modern technology, noting that this encouragement "includes models that use AI."
The EPA justifies this by framing it as a matter of operational common sense. They call predictive modeling a "low-burden, and scientifically sound way to develop inventories without having to perform costly and disruptive excavations at every home." As data scientists, we couldn't agree more. A well-trained model doesn't replace field verification; it optimizes it. It guides your crews with where to dig to get the maximum amount of information with the fewest holes in the street and on private property. Seeing the federal government explicitly recognize AI as a "common-sense" tool is a massive win for forward-thinking utilities.
The Catch: The Primacy Agency Gatekeeper
While the federal EPA is giving a green light to machine learning, they include a critical caveat that every utility must keep in mind: ultimately, acceptance and oversight of its use is determined by your state primacy agency. The EPA provides the overarching encouragement, but your specific state’s department of environmental protection or health holds the final authority on if and how that modeling can be applied and what validation metrics must be met. Navigating the gap between federal encouragement and local state requirements is where the rubber meets the road.
Why You Shouldn’t Wait for the "Final" Version
Because this document was recently released for a brief public comment window, it still carries a "Draft" label. You might be tempted to wait until the EPA publishes the absolute final version before acting.
From our perspective in the data trenches, waiting could mean losing valuable runway.
The core policy direction here is set: the EPA believes in predictive modeling. While they might tweak minor technical phrasing or specific validation metrics in the final text, they are unlikely to reverse course on AI.
Building a reliable machine learning model requires time to audit your existing data, structure your GIS layers, and train the algorithm against your specific historical records. If you wait until the final guidance drops to start your data planning, your utility will be squeezed tightly against the 2027 baseline inventory deadline which defines your replacement target rate.
Let’s Look at the Data Together
At Trinnex, we are data scientists and software developers, but we are water-industry people first. We understand that a model is only as good as the infrastructure data feeding it, and that every water system’s development, data structure, and infrastructure characteristics are unique. Also, every state’s regulatory landscape looks a little bit different.
If you are looking at a mountain of unknown service lines and trying to figure out how to structure a predictive modeling program that your state regulators will actually approve, you don't have to navigate it alone. We’d love to sit down, learn about your specific system challenges, and help you chart a practical, compliant path forward.
Curious about how predictive modeling fits into your state's specific LCRI framework? Reach out to our team today—we’re always down to talk data.
