Ajay Sharma is Lead and Copper Rule data management lead for the New England Area at Kleinfelder. He leads a multidisciplinary team providing support for multiple public water suppliers with LCR compliance.
Nearly 40 years after the Safe Drinking Water Act banned lead pipe installation, millions of Americans are still served by lead service lines. The American Water Works Association estimates that up to 10 million lead service lines remain nationwide, with replacement costs estimated to approach $90 billion. Identifying and replacing buried service lines is a massive logistical and financial challenge facing U.S. cities.
The upcoming Environmental Protection Agency Lead and Copper Rule Improvement deadline of Nov. 1, 2027, adds new urgency. The rule’s intent is to strengthen transparency between utilities and customers, but it also brings complex notification requirements, potential fines and growing public concern at the same time city and water utility staffs are already stretched thin. Before physical replacements can begin, utilities must secure funding, procure contractors and complete field verification — all time-intensive steps.
The good news is that artificial intelligence and machine learning are emerging as powerful tools for managing the daunting workload. These technologies can accelerate inventory development, reduce uncertainty and guide limited resources to where they’re needed most.
The predictive modeling process
A crucial step under the LCRI is reducing the number of service lines in each system that are made of unknown materials. Traditional approaches rely on years of manual reviews and excavations, but modern tools allow even small agencies to build accurate inventories faster and redirect staff to planning and community engagement.
There are two main technology tools agencies with public water systems can use to meet the LCRI’s numerous requirements:
A document intelligence platform, an AI-powered tool built on Microsoft Azure AI Services, significantly accelerates data extraction from existing records.
A predictive model applies machine learning and modern computer science methods to forecast the possibility of lead service lines at any property. This “probability-based” approach, which is approved in most states, allows utilities to support the assertion that remaining unknown lines are unlikely to contain lead. It can be paired with a geographic information system geodatabase to provide a helpful visual representation of lead service lines and property addresses.
Both state and EPA guidance require the use of existing confirmed data to train the model. The model then predicts the probability of lead in service lines with previously unknown materials, giving each service line a probability score ranging from “unlikely lead” to “likely lead.”
With those predictions in hand, engineers and data scientists recommend a targeted inspection list for physical verification through field inspections. The inspection results are fed back into the model to improve model performance. The model increases in accuracy with each iteration, until the probability of remaining service lines containing lead is below a certain threshold.
Real-world success stories
Using AI and ML tools together allows agencies to transform massive volumes of records into actionable insights. For instance, one project processed and digitized approximately 600,000 permits and field documents, followed by 10,000 maps, and merged them into a GIS-based inventory. The customized database aligned directly with EPA and state-specific requirements and integrated seamlessly into a custom built ESRI-based dashboard for visualization and reporting. Reviewing such records manually would have required years of work and extensive staffing. With AI, agencies can now review data in weeks or months, allowing them to focus on decision-making rather than data entry.
Let’s look at two real-world examples of towns that used this approach.
In Yarmouth, Massachusetts, a statistical modeling approach was applied to demonstrate — with 95% confidence — that less than 0.66% of service lines in three districts (South Yarmouth, Yarmouth Port and West Yarmouth) were likely to contain lead or galvanized iron requiring replacement, or GRR, materials. By verifying a randomized inspection pool of 377 service lines and finding no lead or GRR, the town was able to rationalize to the Massachusetts Department of Environmental Protection that a significant portion of its system could be classified as non-lead. This approach helped eliminate thousands of unknowns and allowed the town to reallocate resources to the one remaining area of concern — Hyannis Park — saving time, money and significant staff effort.
In Medway, Massachusetts, the first step included processing a sample set of handwritten tie cards and meter sheets to train the document intelligence platform. The analysis of the remaining cards and sheets quickly indicated the absence of lead, saving considerable time compared with manually reviewing each of the 5,000 documents individually. Resources were then reallocated to enter the DIP outputs into a standardized dataset for efficient critical analysis.
Building public trust
High-profile water crises across the country have eroded public confidence in drinking water systems. Transparent communication and proactive compliance are essential to restore that trust. Using AI-supported tools allows agencies to stay ahead of regulations, minimize risk and communicate progress clearly to residents.
Engaging the community early through transparent updates, clear timelines and open data helps maintain public trust and reduces confusion when construction begins.
AI doesn’t replace human expertise; it empowers staff to do higher-value work. Instead of spending thousands of hours on manual record reviews and data entry, agency personnel can focus on risk management, planning and community engagement. These technologies reduce burnout, improve accuracy and strengthen relationships between utilities and the residents they serve.
Although 2027 may seem distant, cities and towns must act now to avoid falling behind. The longer you wait, the harder compliance will become. By acting now, utilities can reduce unknowns in their inventories, focus field efforts where they matter most, build public confidence through transparency, meet regulatory obligations and save time, money and staff effort.
