Why AI matters at this scale

The East Bay Municipal Utility District (EBMUD) is a public utility providing water and wastewater services to approximately 1.4 million customers in the San Francisco East Bay area. Founded in 1923, it manages a complex system of reservoirs, treatment plants, and thousands of miles of pipeline. As a mid-sized utility in the 1,001-5,000 employee band, EBMUD operates critical, aging infrastructure under significant regulatory and public scrutiny. At this scale, the organization has substantial operational data but faces constraints typical of public-sector entities: tight capital budgets, a retiring skilled workforce, and the imperative to maintain 24/7 reliability. AI presents a lever to transform reactive, schedule-based maintenance into predictive stewardship, optimizing limited resources and future-proofing essential services.

Concrete AI Opportunities with ROI

First, predictive infrastructure analytics offers direct ROI. By applying machine learning to historical break data, soil conditions, and pipe material, EBMUD can create a risk-prioritized capital replacement plan. This shifts spending from high-cost emergency repairs to planned projects, potentially saving millions annually and reducing water loss. Second, treatment process optimization uses AI models on real-time sensor data to dynamically adjust chemical dosing and energy consumption at treatment plants. This ensures consistent water quality while cutting operational expenses, a key metric for rate-setting. Third, intelligent customer engagement via AI chatbots and personalized conservation reports can handle routine inquiries, report leaks via image recognition, and tailor water-saving advice. This improves service while controlling the cost-to-serve, crucial for a large, diverse customer base.

Deployment Risks for a Mid-Sized Utility

Deploying AI at a utility of EBMUD's size involves distinct risks. Legacy system integration is a primary hurdle; data is often locked in decades-old SCADA, GIS, and billing systems, requiring significant middleware investment. Cybersecurity and resilience concerns are paramount, as AI pilots connecting operational technology (OT) to IT networks could create new vulnerabilities for critical infrastructure. Organizational change management is also a steep challenge. Success depends on buy-in from veteran engineers and field crews who trust hands-on experience over algorithmic predictions. A phased, use-case-driven approach that demonstrates quick wins—like predicting pump failures—is essential to build internal credibility and justify broader investment in AI capabilities.