The AI iceberg: Why water utilities are drifting into dangerous waters without deep expertise
In utility boardrooms across the sector, "Artificial Intelligence" has become the mandatory buzzword of every strategic planning session. We are promised that AI will predict pipe bursts before they happen, optimize pumping energy with pinpoint accuracy, and turn our SCADA data into golden insights.
But take a close look at the graphic (ironically generated using AI) above. It illustrates the dangerous disconnect currently threatening the water industry’s digital transformation.
We are building organizations that live entirely on the "Pink Surface." We have populated our utilities with Chief Digital Officers, Innovation Heads, and AI Policy Leads. They stand on the "Strategy & Governance Deck," peering through telescopes at a cloud labeled "Hype." They are talking about AI. They are strategizing about AI. They are governing AI.
But as the illustration shows, they are floating above a massive, submerged reality that they cannot see—and more importantly, that they do not understand.
The illusion of the surface
There is a pervasive assumption among leadership that because they are discussing AI strategy, they possess AI capability. This is the "Pink Box" fallacy.
The people on the boat are essential for setting direction, certainly. But they are often decoupled from the execution. They speak the language of "transformation" and "disruption," but they often lack the vocabulary of "token limits," "vector embeddings," and "training inference."
When an organization is top-heavy with strategists but light on technical talent, they mistake the map for the territory. They see the calm water and the hype cloud, oblivious to the jagged, complex reality beneath the hull.
The 80% reality: Dirty data and hallucinations
Real AI implementation in the water sector isn't about slide decks; it’s about what happens below the waterline. As depicted in the "Blue" section of the graphic, this is where the actual work lives.
AI in water is significantly harder than AI in clean, digital-native industries. Our reality involves:
- Dirty Sensor Data: Noise, drift, and gaps that will confuse any standard model.
- Legacy System Integration: Trying to bolt a Ferrari engine (LLMs) onto a horse cart (SCADA systems from the 1990s).
- Physics-Informed Constraints: A Large Language Model can hallucinate a convincing answer that defies the laws of hydraulics. Without "Physics-Informed Neural Networks" and rigorous validation, AI is just a confident liar.
We need divers, not just captains
The most critical takeaway from this graphic is the staffing disparity. In many utilities, the "Strategy Deck" is crowded, while the underwater teams are nonexistent.
We frequently see utilities attempting to tackle this massive submerged complexity by assigning it to a junior generalist or an enthusiastic intern—the "Jack" of the corporate world. But you cannot send a snorkeler to do deep-sea welding.
To make AI work effectively, we need to value the "Blue" roles as much as the "Pink" ones. We need:
- Senior Water Data Scientists: Professionals who understand both Python and fluid mechanics.
- MLOps Engineers: Experts who can keep models running when data streams change.
- Hydraulic Modeler Liaisons: The translators who ensure the AI respects the physics of the network.
Conclusion
If your organization looks like the top half of this graphic—heavy on strategy, light on technical execution—you don't have an AI capability. You have an AI bureaucracy.
It is time to stop celebrating the breadth of our conversations about AI and start investing in the depth of expertise required to build it. Until the "submerged reality" is managed by true data scientists and engineers, the water sector will remain data-rich, but intelligence-poor.
