The EPA has released two new guidance documents this month, and if you work in service line inventory or replacement planning, both are worth a close read. The August 2026 “LCRI Service Line Inventories – Tips” and “LCRI Service Line Replacement Access – Tips” documents (EPA 815-F-25-005 and EPA 815-F-26-005) update how the agency expects water systems to approach two of the thorniest parts of the Lead and Copper Rule Improvements: identifying unknown service lines and documenting access barriers to replacement.
For utilities leaning on statistical models and AI-based classification to close out their inventories, the inventory tips document is the more consequential of the two. It’s the clearest statement yet from the EPA that predictive modeling isn’t just tolerated — it’s actively encouraged.
Why this update matters right now
The compliance clock hasn’t stopped moving. Systems already completed their initial inventories back in October 2024. The next milestone is the Baseline Inventory, due November 1, 2027, followed by annual updates and a ten-year window — through December 31, 2037 — to resolve any lines still marked “unknown.” That’s a lot of unknowns to close in a compressed timeline, and excavating every unverified service line to check its material isn’t realistic for most systems, cost- or disruption-wise.
That’s the backdrop for why this guidance leans so heavily on evidence-based, non-excavation methods.
The headline: EPA explicitly endorses AI and statistical modeling
The inventory tips document names predictive modeling — including AI-driven approaches — as a legitimate, common-sense way to build out a compliant inventory without inspecting every single service line. Rather than physically checking each unknown line, a system can visually inspect a smaller, randomly selected sample and use that sample as the statistical basis for characterizing the rest of the distribution system, or a defined portion of it.
A few points worth calling out from the guidance:
- It’s tied to a broader federal AI push. The EPA frames this encouragement explicitly within its “Powering the Great American Comeback Initiative,” which aims to position the U.S. as a global leader in artificial intelligence. Water utility compliance is being named as one place where that ambition plays out in practice.
- Neighborhood-level segmentation is sanctioned. The guidance describes dividing a distribution system into sub-areas with shared characteristics — same builder, same era of construction — and running separate statistical models for each. If a sampled subset in a given neighborhood comes back all non-lead, the primacy agency can allow the system to classify the remaining unknowns in that neighborhood the same way.
- A single lead find changes the math. If inspections turn up any lead or partial-lead lines during this process, the system may need to re-segment its modeling approach or fall back on other classification methods for the remaining unknowns in that area.
- Modeling has a role in GRR determinations too. For galvanized lines, a system must show the line was never downstream of a lead service line to avoid classifying it as “galvanized requiring replacement” (GRR). Predictive modeling is called out as one accepted way — alongside historical ordinances or installation records — to make that case for an entire neighborhood, rather than requiring documentation for every single line.
- Early modeling work may reduce future validation burden. Systems that used modeling with rigorous, pre-November 2027 visual inspections — inspections that meet or exceed the LCRI’s own validation standard — may be able to seek a validation waiver, avoiding duplicate effort down the road.
The practical throughline: primacy agency approval is still the gatekeeper, but the EPA is telling agencies and utilities alike that statistical rigor is an acceptable substitute for universal excavation. That’s a meaningful shift in tone from a rule that, on paper, could otherwise push systems toward brute-force verification of every line.
What the access tips document covers
The companion document on service line replacement access is a separate but related piece of the compliance puzzle. It clarifies that:
- Utilities — not the EPA — determine whether they have “access” to conduct a full replacement, and must document it when they don’t.
- Where property owner consent is required, “reasonable effort” is defined concretely: four attempts, using at least two different communication methods, by the applicable deadline.
- Local ordinances, state laws, funding mechanisms, and physical safety hazards can all affect access, and that access can change over time — meaning a system must revisit and replace lines once a barrier clears.
- Worked examples (funding becoming available, ordinances tied to water main projects, state-assigned contractors) show how systems can stay in compliance even with lead and GRR lines still in the ground, as long as access barriers are documented and lines are replaced once access is regained.
The takeaway for utilities
Put the two documents together and the message is consistent: the EPA wants defensible, well-documented decision-making — whether that decision is “this line is non-lead” or “we don’t have access to replace this line yet.” Statistical and AI-based modeling, done transparently and validated appropriately, is presented as a legitimate and encouraged way to get there for inventory classification, not a workaround primacy agencies should be wary of.
For utilities still weighing whether to invest in a modeling-based approach to their unknowns, this guidance is about as clear a green light as the EPA has given. The open question for most systems isn’t whether predictive modeling is acceptable — it’s building the documentation, sampling methodology, and primacy agency relationship needed to use it with confidence.
This post summarizes EPA guidance for informational purposes. It is not a substitute for reviewing the source documents (EPA 815-F-25-005 and EPA 815-F-26-005) or consulting directly with your state primacy agency on your system’s specific compliance approach.
Source Documents:
LCRI Access Tips (pdf)Â
LCRI Service Line Inventory Tips (pdf)
