InfrastructureIssue 036 records
Utility Management
Tonight's utility AI evidence is early and mostly self-reported, but it fills earlier gaps. Houston Public Works is testing an AI surrogate that turns six-hour stormwater model runs into seconds, and UTEP and El Paso Water are starting NASA-funded AI flood mapping that keeps crews in charge. Seville's EMASESA brings 313,000 smart meters and AI anomaly detection onto one platform. SMUD's board report lists AI fire detection, robot substation inspection, next-generation meters and over 100 staff-built GPT models. Maryland co-op SMECO reports quick wins from document automation but no reliability gain yet from outage prediction. A water-sector insider threat report warns that ungoverned AI agents amplify trusted access. US gas and billing AI evidence remains thin.
- Evidence records
- 6
- Cross-source patterns
- 2
- Evidence classes
- 2 vendor claim1 public-sector association guidance1 academic research1 government evaluation1 independent research
- Outcomes
- 4 emerging1 mixed1 cautionary
- Source freshness
- 4 recent2 new this fortnight
- Research completed
- 2026-11-02
Choose a role to see its takeaway beside every record in the ledger.
Patterns across the evidence
Synthesis · Lighthouse Advisory interpretation
Utility AI is being built on consolidated data platforms first
Houston's surrogate flood model depends on a calibrated citywide hydraulic model and data pipelines; EMASESA put 313,000 meters, sector flows and models into one data lake before applying AI; SMUD calls next-generation meters the foundation for distributed intelligence and is integrating OMS with ADMS/DERMS. In each case the data platform is the main build, and AI outcomes are not yet measured.
Operating questionIs the utility's meter, network and model data consolidated and trustworthy enough that an AI tool's output can be checked against it, and who owns that platform?
Supporting evidenceCity of Houston, Houston Public Works (reported by Autodesk)EMASESA (Empresa Metropolitana de Abastecimiento y Saneamiento de Aguas de Sevilla), with Xylem VueSacramento Municipal Utility District (SMUD)
Utilities are keeping AI internal, governed and human-verified while outcomes lag
SMECO reports a time saving from document automation but no reliability gain yet from outage prediction, and keeps member-facing AI on hold under an acceptable use policy. El Paso Water's flood-mapping project keeps crews and emergency procedures in charge and reports confidence scores. The water-sector insider threat report warns that ungoverned AI agents amplify trusted access. Together they argue for scoped, internal, verified uses first.
Operating questionFor each AI use, who verifies outputs before action, what access does the tool hold, and what measured result would justify extending it to customers or operations?
Supporting evidenceSouthern Maryland Electric Cooperative (SMECO), as reported by National Rural Utilities Cooperative Finance Corporation (CFC)University of Texas at El Paso (UTEP) with El Paso Water as research partner; funded by NASAGate15 (Alec Davison, David Pounder); posted on OSF
Evidence ledger
Full record · every source keeps its link and limitations
SMECO turns AI into measurable operational gains: outage prediction, field asset identification and survey automation
SMECO describes three AI use cases. An in-house outage prediction model uses outage history, weather and more than 20 operational variables to forecast device-level failures within a three-day window, is retrained every two weeks and falls back to rules when confidence is low; it has not yet shown measurable SAIDI or SAIFI improvement. A mobile photo-based field asset identification tool linked to GIS and inventory is still a pilot, with benefits stated conditionally. Survey card automation is the clearest return: processing fell from two to three days to one to two hours, an estimated 16 to 24 labor hours saved per cycle. SMECO has an AI acceptable use policy, keeps accountability with the user and has prioritized internal uses before member-facing AI tools.
Why it matters, evidence and limitations
- Why it matters
- A rare co-op account that reports a non-result alongside a time saving. It gives distribution co-ops and small municipal utilities a realistic sequence: start with internal, measurable workflows, treat reliability models as long-horizon work, and set governance before member-facing tools.
- Evidence and measured results
- Self-reported case story published by CFC, a member-owned cooperative lender, quoting SMECO executives. No baseline data, model accuracy figures or independent evaluation are given. The labor-hour saving is an estimate per survey cycle; the reliability outcome is explicitly not yet demonstrated.
- Limitations and uncertainty
- Single co-op, self-reported, no quantified accuracy or financial results; savings are estimates. Published 13 Oct, before this stream's 2026-10-29 and 2026-10-30 editions, and not in the archive; included now because co-op-specific evidence was a recorded gap in both earlier editions and this is the most recent co-op account found that reports a reliability non-result.
Houston Public Works builds a living stormwater master plan, testing an AI surrogate model that cuts model runs from six hours to three seconds
Houston Public Works chose InfoWorks ICM after a documented evaluation of 50 criteria and 20 providers. It built an $11 million citywide Stormwater Master Plan with an interactive public flood-risk map and a planned three-year update cycle. A managing engineer reports that the AI surrogate model returns results in three seconds versus six hours for a full hydraulic run. Autodesk's footnote says surrogate modeling is AI functionality still under development and being tested by customers such as Houston. A separate WEFTEC 2026 session abstract (29 Sep) co-authored by the same Houston engineer describes a wastewater data lake and a hydraulically informed machine-learning model ranking dry-weather sanitary sewer overflow risk to move cleaning from calendar to risk-based schedules, without reporting outcome figures.
Why it matters, evidence and limitations
- Why it matters
- One of the few current U.S. municipal stormwater AI examples: a large city using an AI emulator to make citywide flood modeling fast enough for frequent updates and operational use. Stormwater was a recorded gap in both earlier editions.
- Evidence and measured results
- Vendor-published customer story with direct quotes from the Director of Public Works, City Engineer and a Managing Engineer. The six-hours-to-three-seconds figure is a practitioner statement with no stated accuracy comparison against the full model. The feature is pre-release. The WEFTEC abstract is a conference summary without results.
- Limitations and uncertainty
- Vendor marketing with no independent check; surrogate accuracy versus full model is not reported; the feature is under development. Published 23 Oct, before the 2026-10-29 edition, and not archived; included now because U.S. stormwater AI evidence was missing from both prior editions and this is the most recent U.S. municipal example found.
UTEP and El Paso Water: NASA-funded project to map urban flooding with satellite data and an AI foundation model
A UTEP team received nearly $500,000 from NASA for a two-year project, due June 2028, that combines optical satellite imagery, Sentinel-1 radar and the Prithvi-EO Earth observation foundation model to map flooded areas after storms. El Paso Water is research partner and intended end user. The aim is a dashboard combining flood maps, confidence information and infrastructure overlays so utility staff can prioritize inspections. The team notes that wet pavement, building shadows and swimming pools can resemble floodwater in imagery, and says the tool is not meant to replace crews or emergency procedures.
Why it matters, evidence and limitations
- Why it matters
- A U.S. public water utility partnering with a university on AI for post-storm stormwater and infrastructure assessment, with explicit confidence reporting and human field verification built into the design.
- Evidence and measured results
- University announcement of a funded project; no results yet. Funding amount, data sources, model and timeline are stated. Accuracy targets and evaluation methods are not described.
- Limitations and uncertainty
- Pre-results announcement; benefits are aspirations. Satellite revisit and resolution may limit timeliness for flash floods. Published 8 Oct, before the stream's first edition, and not archived; included now to address the recorded U.S. stormwater gap with a current, utility-partnered project.
EMASESA's Digital Reservoir 5.0 integrates 313,000 smart water meters and AI algorithms on one platform to detect leaks and anomalies
EMASESA's Digital Reservoir 5.0 project, which won IWA Project Innovation Award gold in the Smart Systems and the Digital Water Economy category in early October, brings together 200 network sectors and more than 313,000 smart meters (about 7.5 million readings a day), water quality data, real-time external data, numerical modeling and AI algorithms on a single Xylem Vue platform. The aim is to detect leaks, malfunctions and unusual consumption. EMASESA's head of digital transformation stresses a central data lake and connected rather than isolated solutions. The platform has three layers: a data engine, modules for supply, wastewater and water resources, and operational intelligence for decisions.
Why it matters, evidence and limitations
- Why it matters
- A large public utility's architecture for using AMI-scale meter data with AI, relevant to U.S. municipal water utilities that have deployed AMI but struggle to use the data. Metering and billing analytics was a recorded gap in both prior editions.
- Evidence and measured results
- Vendor-sourced trade press release quoting the utility and vendor. The IWA award is an external recognition, but no measured outcomes (leaks found, water saved, false alarm rates) are reported.
- Limitations and uncertainty
- Vendor-authored, non-U.S., no quantified results. EU data protection and utility ownership structures differ from U.S. municipal settings.
SMUD SD-10 Innovation monitoring report: AI fire detection, robot substation inspection, 100+ custom GPT models and next-generation meters
SMUD's annual board monitoring report on its Innovation strategic direction lists 28 R&D projects in the 2026 report (up from 22 and 25). It describes wide-area AI-enabled fire detection (NextPower Vantage at the Rancho Seco generation site, using AI-based visual processing to spot nearby fires), ANYbotics ANYmal robots paired with substation inspectors for infrared imaging (57 of 270 substations commissioned), staff-built 'SMUD GPT' custom models (over 100, described as increasing internal efficiencies), a new outage management module integrated with existing ADMS/DERMS, and next-generation meters described as the foundation for distributed intelligence applications. Management asked the board to accept the report.
Why it matters, evidence and limitations
- Why it matters
- A public power board's own record of how AI and automation are being governed: reported to elected directors against a written strategic direction. It covers meters, asset inspection, wildfire risk and internal generative AI in one place.
- Evidence and measured results
- Official board exhibit (slides). Claims are management-reported counts and descriptions; no accuracy, cost or reliability outcomes are given for AI tools.
- Limitations and uncertainty
- Slide format with little detail; '100 custom models' is unexplained and unmeasured. Dated 7 Oct, before this stream's first edition, and not archived; included now because it is a primary public-power record covering metering and AI, an area thin in both earlier editions.
Insider Threats in the Water Sector: A Growing and Underestimated Risk (AI agents amplify insider access)
The report argues that insider risk to water and wastewater utilities is rising because of contractor and third-party remote access, cloud services, automation and workforce change. It notes that AI tools and agents without proper governance give faster ways to access, summarize, move or expose sensitive information and operational data, so insider risk now includes how quickly access can be amplified and automated before detection. Mitigations include least privilege and need-to-know access for contractors and vendors, regular access reviews, a scalable insider threat program with defined reporting thresholds, a multidisciplinary threat management group including privacy and HR, and recurring training covering acceptable use.
Why it matters, evidence and limitations
- Why it matters
- Connects AI adoption to the OT and data security risks most relevant to small and mid-size water utilities, where contractors and integrators often hold privileged access.
- Evidence and measured results
- Secondary synthesis by sector practitioners citing WaterISAC, CISA, Ponemon, SentinelOne and a Proofpoint report (87% of organizations use AI beyond pilots, 63% have controls). The AI discussion is brief and relies on vendor survey data; no water-specific AI incidents are documented.
- Limitations and uncertainty
- AI is a small part of a broader insider threat report; statistics are cross-sector and partly vendor-sourced. Dated September 2026 and posted 5 Oct, before this stream's first edition, and not archived; included now as the most recent water-specific source found linking AI agents to OT and insider risk.
How to read this edition
Source findings, measured results and limitations come from the cited publications. Patterns, operating questions, role takeaways and implementation considerations are Lighthouse Advisory interpretation, stated as questions to validate locally rather than guaranteed outcomes. Vendor and operator claims are labeled as claims. Full research method.
- Vendor claim
- A supplier-provided assertion that has not been upgraded to independent evidence.
- Public-sector association guidance
- Practitioner guidance or an association-supplied case; not independent outcome evidence.
- Academic research
- Research produced through an academic institution or peer-reviewed venue.
- Government evaluation
- A public body’s measured evaluation or documented pilot.
- Independent research
- Research conducted outside the implementing organization.