The Resilient Utility: Why AI Is The New Social Safety Net For Water
By Ashwin Dhanasekar and Baker Bozeyeh
In the developed world, water infrastructure is often out of sight and out of mind, until it breaks. But for millions globally, and increasingly in under-resourced communities within wealthier nations, the water system is a daily source of anxiety. Will the pressure hold? Is the water safe? Will the bill be affordable?
As industry practitioners, one from a consulting firm advising major municipalities, the other a founder of an impact-driven water tech provider, we argue that the prevailing narrative around artificial intelligence (AI) in water is too narrow. We often measure AI’s value in return on investment (ROI) or operational efficiency. While valid, this misses a more urgent reality: AI is becoming a critical tool for social equity.
There is also a more uncomfortable truth running underneath the ROI debate: many utilities are not just under-evaluating AI, they are quietly avoiding it. We have spoken with water professionals who describe themselves as “afraid of AI”, wary of the cloud, of automation, and of anything that feels like a black box. While the risks of AI are real and deserve scrutiny, much of the fear is perceptual rather than technical, and it is keeping utilities from tools that could meaningfully improve service, safety, and affordability for the very communities most exposed to infrastructure failure.
For vulnerable communities, the cost of infrastructure fragility is not just financial; it is existential. By shifting our focus from "smart water" to "resilient water," we can use AI to build a social safety net that protects both our infrastructure and the communities that depend on it.
Fragility Is Expensive
The water sector faces a "perfect storm" of challenges: aging infrastructure, climate instability, and a retiring workforce. A 2023 review in Cambridge Prisms: Water highlights that resilience must now be viewed through a "social-ecological-technical" lens, acknowledging that digital hazards and physical failures disproportionately impact marginalized communities (Sinha et al., 2023).
When a main breaks or a pump fails, wealthier districts often have the resilience to absorb the shock. In contrast, under-resourced utilities may lack the budget for redundancy or the staff for rapid manual intervention. The result is prolonged outages or water quality violations that hit vulnerable populations the hardest.
Furthermore, the "graying" of the water workforce creates a dangerous knowledge gap. As senior operators retire, they take decades of institutional memory with them. Without digital capture of this expertise, utilities are left operating via "guesswork," leading to inefficiencies that drive up costs, costs that are ultimately passed down to ratepayers who can least afford them.
Leapfrogging Legacy Technology
The standard roadmap for water utility modernization, installing expensive, on-premises SCADA systems that lock data into local servers, depend on single-vendor stacks, and offer limited interoperability with the modern tools utilities increasingly need, then waiting years to layer on analytics, is no longer the only path. Just as many developing nations "leapfrogged" landlines to adopt mobile phones, water systems can now skip heavy legacy investments in favor of flexible, AI-based tools.
The landline comparison is more than a turn of phrase. Like landlines, SCADA was a remarkable advance in its day and still has a role in many utilities, particularly for basic supervisory control. But it was designed for a world where intelligence lived in a control room, integrations were custom-built one project at a time, and “smart” meant a screen with blinking lights. The expectations for a resilient utility in the age of AI are higher: cloud-native architectures that survive a local outage, interoperable data flows that move freely between SCADA, asset management, hydraulic models, and customer systems, and decision support that does the analysis for the operator rather than dumping raw signals on them. Acknowledging what SCADA does well is not the same as accepting it as the ceiling of what is possible.
Research on "technological leapfrogging" in urban water management suggests that emerging economies and smaller municipalities can transition directly to decentralized, smart systems without the sunk costs of traditional centralized models (Binz et al., 2012).
For example, software platforms that require little or no custom coding allow utilities to deploy advanced leak detection and pressure management without hiring a team of data scientists. By using affordable connected sensors and cloud-based analysis, a small utility can achieve a level of operational sophistication previously thought to be reserved for major metropolitan areas, some of which aren’t necessarily getting their value from SCADA either. The large utilities are often drowning in raw data generated from their old-school monitoring solutions that require them to do the analytics and interpretations manually. Wider access to technology is essential for equity; it ensures that high-quality water service is not the exclusive privilege of the wealthy.
From Passive Dashboards To Agentic AI
To truly build a safety net, we must move beyond passive monitoring. Dashboards that merely display red blinking lights are insufficient if there is no one available to interpret them. The future lies in "Agentic AI", systems capable of not just detecting anomalies but autonomously acting to resolve them.
Consider a scenario where a sudden pressure spike threatens a fragile pipe network in a low-income neighborhood. A traditional system might send an alarm to an overwhelmed operator who sees it hours later. An Agentic AI system, however, could detect the event in real time and trigger the appropriate protective response: isolating the affected zone, alerting an on-call crew, or engaging pre-engineered surge protection such as air vessels or relief valves. The distinction matters: genuine hydraulic transients propagate at acoustic wave speeds, on the order of milliseconds, faster than any control valve can actuate, and an abrupt valve movement can itself induce a damaging surge (Ghidaoui et al., 2005). Mitigating water hammer is therefore a matter of detecting the event and engaging purpose-built protection, not throttling the wave with a valve in real time.
It is worth pausing on the word “autonomous”, because that is the part that makes utilities most uncomfortable. Agentic AI does not mean unsupervised AI. In well-designed deployments, the operator stays in the loop for any consequential action: the system recommends, escalates, or stages a change, and a human approves it before it propagates. AI handles the speed and the scale; people retain the judgment and the authority. Framed this way, “autonomy” is less about taking humans out of the loop and more about freeing them from the noise so they can focus on the decisions that actually require them.
Utilities also do not need to leap straight from a static dashboard to a fully agentic system. There is a useful middle layer: AI assistants and chatbots that sit on top of existing data and let an operator ask, in plain English, “why did flow in Zone 4 drop overnight?” and receive a synthesized answer drawn from SCADA, hydraulic models, and work-order history. Forecasting tools, anomaly detection, and automated image analysis of CCTV pipe footage offer similar incremental wins. These intermediate steps are easier to introduce internally, give staff a chance to build trust in the technology, and create the muscle memory that makes the eventual step toward agentic control far less daunting.
This capability is supported by recent findings on digital resilience, which define it as the ability of information systems to not just detect shocks but "absorb major stresses" and recover service levels (Sinha et al., 2023). By automating these routine "health checks," we free up human operators to focus on complex, high-value tasks that require empathy and judgment, such as customer service and community engagement.
Workforce As Stewardship
Critics often fear that AI will displace jobs. However, in the context of a labor shortage, AI is not a replacement; it helps a limited workforce do more. A study on the water workforce emphasizes that embracing AI is crucial to attracting a new generation of talent that grew up with digital tools and retaining institutional knowledge (Smart Cities Dive, 2025).
As we develop curriculums for the next generation of environmental engineers, work currently underway at institutions like Colorado State University, we are not just teaching code. We are teaching stewardship. The "Engineer of 2030" must understand business process modeling and data science, but they must also understand their role as custodians of public health. AI tools provide the "digital mentorship" that allows younger engineers to make safe, informed decisions earlier in their careers.
Conclusion
We cannot solve 21st-century water problems with 20th-century tools. If we treat AI solely as a luxury for efficiency, we widen the gap between the water-rich and the water-poor. But if we deploy it as a tool for resilience helping with enabling leapfrogging, automating protection, and empowering the workforce, we can ensure that reliable, affordable water is a guaranteed right for every community.
Concretely, the resilient utility we are describing looks like this:
- Cloud-hosted analytics that keep working when a local server fails
- Sensor coverage dense enough to catch a leak before a customer does
- A data layer that lets SCADA, asset management, hydraulic models, and customer billing speak the same language
- AI that flags pressure transients in seconds and tells the on-call operator what action to take, with the option to approve or override
- A workforce comfortable querying the system in plain English
Getting there is rarely a single, multi-layer megaproject. In our experience, it usually starts with a focused pilot (non-revenue water, leak detection, or pressure management are common entry points) and grows in scope as confidence and capability compound. The goal is not to digitize everything at once, but to put a credible first win on the board and build from there.
What that first win actually requires is less exotic than the technology around it suggests. Most utilities already have the raw material: years of SCADA history, billing records, a GIS that is at least roughly accurate, and metering at the district level. The early work is not necessarily buying hardware; it is making that existing data trustworthy and putting it somewhere it can be cross-checked. A utility that cannot yet close its own water balance will struggle to act on an algorithm’s output, however sophisticated the algorithm.
After that, the order of operations matters more than the choice of tools. Fix what happens when an alert fires (who receives it, how fast they can reach the asset, and whether time-to-repair is measured) before investing in cleverer ways to generate alerts. Bring the operators who will live with the system into it early, and give them the authority to override it; a tool only earns its place once the people using it trust it, and that trust comes from training and hands-on control. Make interoperability a condition of purchase rather than a hoped-for bonus: ask not only what a system does on its own, but whether it will still share its data with your SCADA, hydraulic model, and billing system years from now. Each of these steps earns the next, which is why the utilities that get there tend to have started small and stayed patient.
Water justice in the digital age requires silicon as much as it requires steel. It is time to build the digital infrastructure that serves as a true social safety net.
References:
- Binz, C., Truffer, B., Li, L., Shi, Y., & Lu, Y. (2012). "Conceptualizing Leapfrogging with Spatially Multi-scalar Innovation Systems: The Case of Onsite Wastewater Treatment in China." Technological Forecasting and Social Change.
- Ghidaoui, M. S., Zhao, M., McInnis, D. A., & Axworthy, D. H. (2005). "A Review of Water Hammer Theory and Practice." Applied Mechanics Reviews, 58(1), 49–76.
- Sinha, S. K., et al. (2023). "Water sector infrastructure systems resilience: A social–ecological–technical system-of-systems and whole-life approach." Cambridge Prisms: Water, 1-20.
- US Water Alliance. (2023). "Advancing Water Equity in Small and Rural Communities: The Role of Digital Solutions."
- Various Authors. (2025). "The water workforce is graying. Here's what we need to do to restart." Smart Cities Dive.
Ashwin Dhanasekar is the Operations Leader for Digital Solutions at Brown and Caldwell, based out of Denver, CO. He also leads BC’s AI and Data Science Team, helping the water sector organically and effectively adopt AI. Ashwin is also an Adjunct Professor at Colorado State University, helping students apply AI in practice. He has over 15 years of experience in the water sector, working in various roles throughout the sector, helping clients with their digital transformation.
Baker Bozeyeh is director of Flowless. Flowless is on a mission to help water utilities in improving their daily operations through AI-aided analytics and control automation. Flowless software app guides leak managers and network operators through all steps needed to detect and reduce water losses and optimize operations, with AI-powered insights generated and explained by Flowless AI agent.