Request Demo

Case Study

Water Utilities

6x ROI and 75% Faster Response: How the St. Louis Water Division Modernized Infrastructure with Sand and AWS
A real-time digital twin and AI system that transformed reactive water main break response into proactive, data-driven infrastructure management.
Return on investment
0 x
Faster boil water advisory issuance
0 %
Annual overhead reallocated to strategic projects
$ 0 K
Of aging infrastructure now continuously monitored
~ 0 mi

Executive Summary

Sand deployed a production-grade digital twin and AI system on AWS to modernize the St. Louis Water Division’s reactive response to over 300 annual water main breaks. By integrating siloed SCADA and GIS data using services like AWS DMS, Amazon Redshift, and AWS IoT Core, the solution automates burst localization and traces downstream impacts in real time. This high-impact architecture delivered a 6x financial return on investment, slashing boil water advisory timelines by 75% and allowing the utility to reallocate $300,000 in annual management overhead directly into strategic projects.

Problem Statement

The St. Louis Water Division manages over 1,400 miles of aging water infrastructure serving more than 300,000 residents, a system that experiences upwards of 300 water main breaks each year. Historically, the utility’s leak detection and response workflow was highly reactive, relying on customer complaints and manual investigations that left operational teams without a unified, real-time view of the distribution network. When a burst occurred, operators lacked the system intelligence to automatically isolate the break or identify affected downstream customers. Instead, executive leadership was forced to manually map out impacted areas using Google Maps, a highly probabilistic, error-prone process that cost the utility approximately $300,000 annually in diverted leadership hours. Because operational data was siloed across disparate SCADA, GIS, and work order systems, pinpointing failures and issuing mandatory boil water advisories routinely took over four hours. This severe delay not only prolonged service disruptions but also significantly increased the risk of public exposure to water contamination.

Solution Deployed

To address these fragmented operations, Sand engineered a production-grade, standalone digital twin system deployed within a secure AWS environment, breaking down data silos across customer service, control rooms, repair crews, and leadership. At the core of the data integration layer, AWS DMS (Database Migration Service) seamlessly consolidated disparate datasets from SCADA, GIS, and legacy work order systems into Amazon Redshift and Amazon RDS Postgres, creating a single, high-performance data warehouse for cross-departmental intelligence. To solve the critical challenge associated with low water pressures due to bursts, Sand deployed an application built on Amazon ECS and EC2 that provides real-time sensor telemetry via AWS IoT Core. These models analyze pipe-to-SCADA associations, live pressures, and historical asset registers to instantly identify burst events and track the exact valves that need closing to isolate the burst.

By mapping these complex, interconnected water network infrastructure into Neo4j AuraDB, the system instantly traces downstream dependencies, while Amazon ElastiCache (Redis) ensures the ultra-low latency required for real-time spatial queries. When a break occurs, the system utilizes AWS Lambda functions that integrate with Amazon SNS/SQS to run calculations that determine the areas affected by low water pressures. This boundary data is then used to synchronize multi-departmental response in near-real-time:

  • Customer service teams are immediately alerted to notify impacted residents via automated workflows
  • Maintenance crews receive precise dispatch coordinates
  • Leadership can instantly anchor boil water advisories
These changes slashed the previous four-hour delay to mere minutes. Secured by Amazon Cognito for role-based access control and optimized globally via Amazon CloudFront, this architecture transforms the St. Louis Water Division from a reactive utility into a proactive, data-driven operation capable of safeguarding public health and strategically prioritizing capital expenditure infrastructure investments.

Outcomes and Success Metrics

The deployment of the AWS-backed AI system has delivered immediate, measurable improvements in public safety, operational efficiency, and financial stewardship for the St. Louis Water Division, yielding an extraordinary 6x return on the utility’s initial financial investment. Most notably, the utility achieved a 75% efficiency gain in issuing boil water advisories to affected residents. A process that previously took hours, and sometimes days, of manual mapping and cross-departmental coordination now occurs near-instantly, allowing accurate, localized safety alerts to be deployed to the public within minutes of a confirmed infrastructure failure.

Furthermore, by automating the network isolation and customer impact mapping workflows, the utility completely eliminated the manual Google Maps process, successfully reallocating $300,000 back into strategic activities, including long-term funding strategies, asset reliability and pipe replacement modeling, and revenue-generation initiatives. The success of this phase has served as a proof of concept for broader digital transformation, triggering the adoption of additional advanced use cases, including the planned construction of a centralized, real-time Intelligence Center to give the Head of the St. Louis Water Division, the St. Louis Mayor, and the city’s COO a unified operational dashboard for citywide infrastructure.

Lessons Learned

The successful deployment at the St. Louis Water Division highlighted three critical pillars for executing high-impact AI projects in the public sector.

  • Active leadership and executive buy-in proved pivotal from day one; having champion stakeholders ensured the project targeted the right operational friction points, established clear ownership, and paved a smooth path for cross-departmental adoption.
  • Partnering with domain-specific AI experts was critical to accelerating time-to-value, bypassing the steep learning curve of building complex hydraulic and machine learning physics models from scratch.
  • Prioritizing explainable and responsible AI was essential for building organizational trust. By ensuring that operators and engineers could clearly understand why the system recommended specific valve isolations or flagged specific pipes at risk, the utility fostered deep confidence in the technology, shifting the organizational culture from skepticism to enthusiastic, widespread adoption.

Get started

Physical AI for Critical Infrastructure.

Request Demo