What AI and Smart Metering Actually Do to Your Books

Advanced metering infrastructure, leak-detection analytics, and AI-driven demand forecasting have moved from pilot programs to line items in a lot of water utility capital budgets this year. The technology pitch is usually about non-revenue water reduction and operational efficiency, and that pitch is often accurate. What gets less attention in the sales conversation is how these projects actually flow through the accounting, and that's where I see finance staff get surprised months after the board approved the project.

Not Everything in an AMI Project Is Capital

Advanced metering infrastructure — meters, endpoints, the communication network, and the head-end software that receives the data — is the easy part of this determination. The physical meters and network equipment are capitalizable, depreciated over their useful life the same way any other distribution system asset would be. That much hasn't changed from how utilities have always treated meter replacement programs.

Where it gets more complicated is the software and analytics layer sitting on top of the metering data — the platform that turns fifteen-minute interval reads into leak alerts, consumption anomaly detection, or AI-driven demand forecasts. A lot of these platforms are sold as software-as-a-service, billed monthly or annually rather than through a one-time license purchase. That distinction matters enormously for how the cost gets recorded.

A rough map of what falls where

Cost component	Typical treatment
Meters, endpoints, communication network hardware	Capital asset — depreciable distribution plant
Head-end system you own and operate on your own servers	Capitalizable internal-use software, amortized over useful life
Vendor-hosted SaaS analytics/AI platform (subscription)	Subscription asset/liability under GASB 96, or expensed as incurred outside GASB, depending on implementation-cost criteria
Ongoing data hosting and platform fees	Operating expense, typically
Installation, testing, data conversion for owned systems	Capitalizable if directly attributable to putting the asset into service

I've seen finance staff default to capitalizing the entire AMI project cost, including the ongoing subscription fee for the analytics platform, because it was bundled into one capital budget line by the operations department that requested it. That's the wrong instinct. The capital budget approval and the accounting treatment are two different questions, and a board approving a project as "capital" doesn't make every dollar in it capitalizable under GAAP or GASB standards.

Where the Non-Revenue Water Savings Actually Land

The business case for AMI and leak-detection AI is almost always built around reducing non-revenue water — treated water that's produced but never billed, whether from leaks, meter inaccuracy, or unauthorized use. That's a real and often substantial benefit, but it doesn't show up on the income statement as a line item labeled "savings." It shows up as reduced production and treatment costs (lower chemical and power expense for the volume you no longer have to produce and lose), improved billed consumption relative to production, and in some cases the ability to defer a treatment plant capacity expansion because you're no longer producing water that leaks out before it's billed.

That last one is worth pausing on. If AMI-driven leak reduction genuinely defers or eliminates the need for a capacity expansion project, that's a real capital avoidance — but it needs to be tracked and documented as such if you want credit for it in a future rate case or capital plan review. "We didn't have to build the plant expansion because non-revenue water dropped from 22% to 14%" is a much stronger rate case narrative than letting that benefit disappear into a general improvement in the numbers nobody can trace back to the AMI investment.

Depreciation Life Is a Real Planning Variable, Not an Afterthought

Smart meters and the associated communication network have a materially shorter useful life than the pipe and treatment infrastructure utilities are used to depreciating over 40-50 years. Endpoint technology in this space is still evolving quickly, and a 15-20 year useful life assumption is more realistic for AMI equipment than the longer schedules applied to civil infrastructure. Understating that useful life difference in a capital plan means underfunding the eventual technology refresh, and a utility that treated its first AMI rollout as a one-time capital event is often unpleasantly surprised when the network needs meaningful reinvestment a decade and a half later.