In Martin County, Florida, the problem began with a blank. A water utility serving more than 100,000 residents had 26,338 service lines whose materials were unknown. The pipes existed. Water moved through them. The records could not reliably say what was underground. For an organization trying to account for lead in its network, that was an expensive kind of ignorance.
- Trinnex combines water engineering, cloud software, and data science.
- Its leadCAST tools organize pipe inventories and help target inspections.
- The business extends into sewer forecasting, water quality, and capital planning.
The obvious response would be to send crews out to inspect. Repeat that thousands of times and an information problem becomes a construction budget. Trinnex, the software and consulting subsidiary of CDM Smith, offered another sequence: consolidate the records, inspect a selected group of lines, and use statistical analysis to make the remaining uncertainty more manageable.
852 places to start
Working with CDM Smith and Trinnex, Martin County brought service line inventory, billing records, and geographic information into leadCAST. Crews then verified 852 recommended lines using its field application. Trinnex documented the statistical approach for the county’s submission to Florida’s Department of Environmental Protection. More than 26,000 lines were subsequently classified through that approach without inspecting each one.
field verifications informed a much larger classification effort.
The company’s case study puts the avoided inspection cost at approximately $20 million. That is a reported estimate of work avoided, rather than an audited return on software spending. It also concerns classification, rather than thousands of pipes physically replaced. Those distinctions give the number its meaning: better information can change the amount of fieldwork a utility needs to commission.
There is a useful wrinkle in the case study’s terminology. It calls the resulting category “unknown non-lead.” A model-supported classification and a directly observed pipe are different kinds of evidence. Keeping track of that difference is part of the work, particularly when a regulator must understand how the inventory was assembled.
A startup with engineering in its pockets
CDM Smith launched Trinnex in January 2022 with Amy Corriveau as president. The parent supplied something a new software company usually has to acquire slowly: familiarity with the infrastructure owner’s problems. Its initial products addressed sewer performance, lead service lines, rainfall, and wastewater surveillance. The name itself joined three ideas: trust, innovation, and experience.
That inheritance explains the product’s shape. leadCAST joins inventories, sampling results, field records, replacement progress, and resident communications. The user may be an engineer at a desk, a crew member holding a phone, or a program manager answering a homeowner. A technically impressive prediction has limited value if nobody can connect it to the next inspection or explain it to the public.
“What sets leadCAST apart is their team.”Anne M. Murray, Martin County Utilities, in Trinnex’s case study
There are alternatives. Water-quality software provider 120Water and predictive-modeling specialist BlueConduit announced a partnership in 2023. Trinnex therefore competes on more than the presence of machine learning. Its proposition combines engineering advice, program administration, predictive modeling, and implementation support. In May 2025, it connected leadCAST Predict with Esri’s ArcGIS inventories, allowing those customers to work from a central dataset.
The invoice is part of the story
Trinnex earns money through subscriptions, implementation, and consulting. A May 2024 proposal in Tamarac, Florida’s public meeting materials makes that arrangement concrete. It quoted $33,250 for leadCAST Tier 2 implementation, $9,180 annually for the core subscription, and $29,000 for the first year of leadCAST Predict. Predict’s quoted annual price declined in the following two years.
Customer-specific quoted prices; separate items, different billing periods.
The full three-year proposal, including website services, totaled $154,460. This gives buyers a useful budgeting question: what portion purchases access, and what portion pays for getting the organization ready to use it? Implementation included configuration, data ingestion, and introductory training. Predict included model development, evaluation, reporting, and consulting. Comparing only subscription prices would miss much of the assignment.
When the sewer disagrees with the model
The same attention to evidence runs through waterCAST Sewer. It combines existing utility systems and compares measured conditions with hydraulic expectations. A mismatch can prompt investigation of a blockage, an overflow, or a data issue. In New Bedford, Massachusetts, the published case describes a dashboard combining rainfall, overflow depths, simulations, and forecasts, plus automated overflow notifications.
Other tools move the decision further upstream. Capacity Planner helps assess sewer permit requests and track reserved capacity. waterCAST WQ centralizes water-quality information and reporting workflows. assetCAST supports risk-based replacement priorities and capital plans. Consulting covers GIS, cybersecurity, data integration, digital twins, and custom software. Together, they occupy the space between collecting infrastructure data and committing public money.
Give the next person something usable
In January 2026, Trinnex integrated more than 100 GIS and asset management professionals from CDM Smith. In June, it launched Raini, an AI agent ecosystem designed to connect operating procedures with utility data. Its stated targets include slow information retrieval and knowledge lost when experienced workers retire. The expansion makes the human side of its software business hard to miss.
Security belongs in that discussion, too. Trinnex completed a SOC 2 Type 2 audit in September 2024, with A-LIGN as the assessor. For buyers connecting operational records and public-facing workflows, scrutiny of the supplier sits alongside scrutiny of the model. A useful dashboard still has to earn permission to handle the information behind it.
For readers managing other assets, the copyable practice is straightforward: assemble the records, collect deliberate observations, and explain how the recommendation was reached. The conditions matter. Weak records need cleaning; predictions need local testing; regulators need suitable documentation. Trinnex itself says there is no universal model for predicting pipe material. The useful promise is a better-supported decision about where the next crew should go.