Water systems are vast, interconnected, and critically important. Planning and managing these huge networks requires highly-specialized engineers and sophisticated tools, like hydraulic models and, more recently, digital twins. Understanding how a complex network will react to a sudden pipe burst or even just routine maintenance requires specialized expertise.
That reliance on deep expertise, which isn’t always available, is often a bottleneck. This is where the emerging field of agentic AI for water utilities comes into play.
Agentic AI is a type of artificial intelligence that can make decisions and take actions on its own to achieve a specific goal. Think of it not as a passive tool but as a smart co-pilot. But this new capability leads to a crucial question, one we recently posed to the water community on LinkedIn: As utilities explore AI for efficiency and resilience, how far should we go in giving it control?
The results from 174 votes were crystal clear:
The overwhelming consensus (a combined 90%) points toward collaboration, not replacement. The industry appears optimistic but cautious. This “human-in-the-loop” model, where 55% of professionals are comfortable with AI agents making recommendations for human approval, seems to be the industry’s preferred path for exploring agentic AI in water operations.
How Agentic AI and Water Systems Can (Safely) Work Together
When many people think of AI, they imagine a single, all-knowing system—”generic water ChatGPT” expected to do everything. But in reality, managing critical infrastructure may be far too complex for a one-size-fits-all solution. What’s more, working with “generic LLMs” or feeding them with generic prompts means a greater likelihood of “hallucinations”.
The agentic AI approach proposes a different, more manageable structure.
Think of it as building a dedicated, expert team rather than hiring one overwhelmed generalist. Each AI agent is a specialist, designed with a clear, specific goal. This focus could be its superpower.
For example, a utility’s “digital team” might include:
A Data Agent to gather real-time data from sensors and meters.
A Simulating Agent to run complex hydraulic model “what-if” scenarios.
A Coding Agent to help integrate new data sources or platforms into the utility’s workflow and produce visualisations.
An Orchestrating Agent to coordinate the other agents and present a single, actionable insight to the human operator for approval.
This collaborative structure would allow the AI to perform complex calculations, but deliver a clear, simple recommendation, keeping the human operator firmly in control.
AI Agents in Action: A Utility’s New Digital Team
So, what would this “team” of AI agents actually do? Here are a few practical examples of how agentic AI and water management could intersect:
The Operator’s Co-Pilot (Digital Twin Agent): This agent runs hydraulic models to observe differences between observed and computed values, and if there is a significant discrepancy, it can raise an alarm and initiate a remedy. It could work 24/7, for example monitoring data to detect the subtle patterns of a new leak, checking previous reports of leaks in that area, and indicating corrective measures, potentially hours before it becomes obvious. It could also run thousands of simulations to answer questions like, “What’s the most energy-efficient pump schedule for tonight, given the forecast?”
The Customer Service Agent: This agent could handle the front lines of customer interaction, instantly answering common billing questions or providing updates on service interruptions. This would free up the human team to manage the most complex and sensitive customer relationships.
The Regulatory Research Assistant: This agent could act as a dedicated research assistant. An operator might ask it, “Summarise the latest reporting requirements for water quality,” and it could scan and synthesise the necessary documents, presenting a clear summary for human review.
The Talent & Training Agent: The “silver tsunami” is a major challenge. This agent could help bridge the gap by assisting HR in identifying qualified candidates, pinpointing skill gaps in new hires, and suggesting specific training modules or standard operating procedures (SOPs).
The Future is Collaborative, Not Autonomous
If that 55% majority in our poll is any indication, that is those prepared to grant AI agents control, subject to human approval, the industry sees a clear and practical path forward. The conversation about agentic AI and water isn’t about a “risky takeover” or “fully autonomous agents.”
Instead, it’s about exploring the potential of reliable, focused AI co-pilots that empower, assist, and inform human experts. This approach could dismantle the expertise bottleneck and make advanced digital water capabilities more accessible. It’s not about replacing the operator; it’s about giving them better tools.
By focusing on smart, manageable agents that work with operators, we can begin to explore how to build a more resilient, efficient, and collaborative water future for everyone.
