What challenges did EPCOR face before AI implementation?

High levels of undetected water leaks led to non-revenue losses of up to 40% for EPCOR and other utilities globally, with some regions losing as much as 70% of piped drinking water. Traditional leak detection methods cannot reliably detect leaks, especially in plastic pipes, leading to long repair cycles and unnecessary excavation costs. Growing population and rapid service area expansion put added pressure on water infrastructure in arid and drought-prone regions. Lost water increases costs related to pumping, treatment, and energy—costs ultimately paid by customers. Utilities are expected to uphold sustainability goals and better manage limited water resources.

What was the impact of the AI solution?

EPCOR reduced non-revenue water from 27% to 10% in its Arizona network, decreasing water loss and saving millions in treatment and pumping costs. Improved repair prioritization enables targeted interventions and quicker fixes, reducing street and property disruptions. Utilities using the solution saw drastically faster leak detection and repair timelines. Supports large-scale water conservation and sustainability initiatives in drought-sensitive regions through more efficient operations.

Source

MicrosoftProductionEvidence: Medium65/100

EPCOR Cuts Water Waste via AI Leak Detection

Use case typeIrrigation optimizationUpdated Jun 13, 2026

EPCOR, Arizona's largest private water utility, faced the problem of high water loss—up to 40%—due to undetected leaks in its vast piped network, a challenge shared globally where some regions lose as much as 70% of their drinking water. To address this, EPCOR partnered with UK-based FIDO Tech and Microsoft. FIDO's AI acoustic tool, enhanced by GPT-4 and deployed on Azure OpenAI Service, analyzes acoustic sensor data from water pipes to precisely locate leaks, rank them by severity, and prioritize repairs. Since 2023, this technology has been deployed in several geographies, including Arizona, Mexico, South Africa, and the UK, in collaboration with organizations such as State Water Commission in Querétaro and Las Vegas Valley Water District. Microsoft's global water sustainability initiative leverages such AI-driven projects to support watershed replenishment and operational efficiency across utilities. The AI tool's ability to accurately detect even quiet leaks in plastic pipes, which have traditionally challenged old detection methods, leads to much faster repairs and reduced excavation costs for communities. FIDO sensors are attached to pipe infrastructure, capturing acoustic data that is then processed by deep learning models to differentiate genuine water leaks from background noise and other non-leak sources. The system can recommend optimal sensor placement, assess water volumes lost, and help schedule repairs for maximum impact. This advanced solution streamlines leak detection, lowers multi-million-dollar pumping and treatment costs, and supports water conservation efforts vital to arid regions. EPCOR's implementation resulted in a drop of non-revenue water from 27% to about 10%, translating into major cost savings and reduced supply strain. Due to the scalability and flexibility of FIDO's solution, sensors can be redeployed and moved, providing lasting benefits across rapidly expanding service areas. This project is a model for combining advanced AI, IoT, and deep learning to address pressing environmental and infrastructural challenges, demonstrating transferable benefits for utilities worldwide.

Organization: EPCOR
Industry: Energy & Utilities
Location: United States
Published: September 2024

Reported outcomes

10-27%

costCost savings

Strategic outcomes

Cost efficiencyReduced non-revenue water lossesSpeed & agilityAccelerated leak detection and repairCustomer experience & trustReduced street and property disruptionSustainability & ESGImproved water conservation efforts

Catalog median for cost savings deployments: −40% across 177 reported metrics. Compare benchmarks →

Why do we believe this?Outcome claims, sources, and evidence checks

Normalized claim

Cost: 10-27% decrease

news.microsoft.comSep 19, 2024News articleInferred claimMedium evidence strength

EPCOR reduced non-revenue water from 27% to 10% in its Arizona network, decreasing water loss and saving millions in treatment and pumping costs.

Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks

Customer: EPCOR
Provider: Microsoft
Maturity: Production
Linked source: news.microsoft.com

FIDO's AI acoustic tool, enhanced by GPT-4 and deployed on Azure OpenAI Service, analyzes acoustic sensor data from water pipes to precisely locate leaks, rank them by severity, and prioritize repairs

Customer identity supportedSource describes one deploymentMaturity supported

Primary read

Use case focus

Showing 3 of 3

1AI-Driven Acoustic Leak Detection in Municipal Water Systems
2Non-Revenue Water Reduction for Utilities
3Automated Leak Localization and Prioritization

High levels of undetected water leaks led to non-revenue losses of up to 40% for EPCOR and other utilities globally, with some regions losing as much as 70% of piped drinking water.
Traditional leak detection methods cannot reliably detect leaks, especially in plastic pipes, leading to long repair cycles and unnecessary excavation costs.
Growing population and rapid service area expansion put added pressure on water infrastructure in arid and drought-prone regions.
Lost water increases costs related to pumping, treatment, and energy—costs ultimately paid by customers.
Utilities are expected to uphold sustainability goals and better manage limited water resources.

Deployed FIDO Tech's AI acoustic leak detection system, integrating deep learning models powered by GPT-4 via Microsoft's Azure OpenAI Service.
Sensors attached to water pipes collect acoustic data, which is processed by AI models to pinpoint and prioritize leaks.
Solution works for different pipe materials, including difficult-to-monitor plastic pipes, and provides real-time recommendations for sensor placement and post-repair validation.
Microsoft fostered deployment through a global water replenishment initiative and local collaborations.
The system interfaces naturally with field technicians using conversational AI.

Technologies

Azure OpenAI Service GPT-4

EPCOR reduced non-revenue water from 27% to 10% in its Arizona network, decreasing water loss and saving millions in treatment and pumping costs.
Improved repair prioritization enables targeted interventions and quicker fixes, reducing street and property disruptions.
Utilities using the solution saw drastically faster leak detection and repair timelines.
Supports large-scale water conservation and sustainability initiatives in drought-sensitive regions through more efficient operations.

Architecture

Sensors deployed on water pipes record acoustic signals. Data is transmitted to the cloud, where GPT-4 models running on Azure OpenAI Service analyze the signals for leak signatures. The system ranks leak severity and advises utilities on repair priority and optimized sensor placement. Conversational AI interfaces enable technicians to interact with the platform and validate repairs. Real-time cloud analytics enable rapid scaling across global service areas.

Implementation partners1

FIDO Tech

Sources & evidence1

Evidence: Medium65/100Evidence strength