Garbage In, Insight Out: How AI Is Cleaning Up Water's Data Problem

By Christian Bonawandt

The world’s water and wastewater infrastructure is at the crossroads of several critical challenges: aging assets, mounting climate pressure, increased industrial demand, and a severe shortage of skilled engineering professionals. While the industry has the talent and technology to overcome each of these, they are often hamstrung by inconsistent or unreliable data. At the Bentley Year in Infrastructure 2025 Conference in Amsterdam (Oct. 14 to 16), industry leaders discussed how artificial intelligence (AI) is transforming this fragmented data landscape, enabling utilities to move from reactionary maintenance to proactive, adaptive planning.

Dealing With Legacy Data

Because much of the global water sector’s infrastructure was constructed decades ago, critical data is trapped in silos, outdated formats, or exists only as handwritten notes that is often institutional knowledge of experts nearing retirement. As Nuno Medeiros, head of asset management at EPAL (Water Supply Company of Lisbon), pointed out, many utilities that adopted early technologies like SCADA and geographic information systems now grapple with a new problem: they have more and more data from sensors but are lacking the “system that can extract information from them” and integrate it with established data sources. Richard Vestner, VP of cities and water for Bentley Systems noted that historical performance information is often “completely broken” and difficult to tie back to original design.

According to Rod Naylor, global water lead at engineering firm GHD, this data deficit is especially risky because water infrastructure is fundamentally “hyper local”. Unlike virtual energy trading, water management is physical, running through pipes and constrained by local regulations, state laws, and even municipal mandates. During opening remarks at the conference, Nicholas Cumins, CEO of Bentley Systems, affirmed the high stakes: “Engineers work in a creative profession but one where precision is not negotiable and consequences are real.”

Digitizing The Unseen And Validating Quality

AI solutions are directly addressing the “garbage in, garbage out” problem by creating structured context from chaos and ensuring that new operational data is grounded in engineering reality. Some examples include:

1. Modeling Aging Assets: For massive legacy assets like dams, which many regions are focused on rehabilitating rather than replacing, AI helps create the necessary digital foundation. By leveraging reality capture capacities, Bentley’s tools can help utilities digitize legacy information, such as converting existing 2D digital plans into 3D models (scan to building information models/BIM).
2. Automated Anomaly Detection: AI vision capabilities can examine these digital models to automatically detect physical issues like cracking or spalling (i.e., surface failure), immediately flagging these anomalies and creating insights about the asset's condition.
3. Connecting Operational Data to Physics: For new data streams, such as those from increasingly deployed sensors, AI ensures relevance. Cumins stressed that AI is leveraged to connect IoT data to engineering principles, allowing operators to understand why the data matters and determine if a specific data series is affecting the asset's functional requirement.
4. Specialized Validation: AI-powered systems are being paired with established engineering applications to ensure trustworthiness. In Brazil, for example, a leak-detection system powered by AI was checked against hydraulic models to verify accuracy, resulting in a one-third reduction in water validation complaints and the recovery of over 175,000 cubic meters of water monthly. Similarly, engineers have used AI to calibrate subsurface geotechnical models — data critical for foundation work in water structures — demonstrating the necessity of engineering context.

Accelerating Adaptive Planning

AI is also enabling the water industry to transition from slow, conventional planning to the rapid, adaptive strategies needed for managing the uncertainty driven by climate change. Naylor explained that adaptive planning involves creating alternative pathways and identifying trigger points for investment. This approach, enabled by digital technologies, accelerates decision-making from 10-year planning cycles to months or, in some cases, weeks.

The key to this speed is building institutional memory via a digital thread, which connects design, construction, and operations. As Cumins explained, by training AI assistants on decades of product documents, standards and expert insights, engineers gain immediate access to knowledge, empowering even junior team members.

This institutional data then fuels sophisticated AI tools like the Bentley Copilot, an infrastructure AI assistant integrated directly into products. This tool is trained on civil engineering concepts and principles and can handle tasks like calculating complex hydraulics based on storm data, interrogating requirement documents, and making complex design changes based on verbal input.

Speaking exclusively to Water Online, Vestner noted that this same AI capability can review local regulations and laws that have been uploaded into its database to ensure regulatory compliance and validate standards. “The machine does it for you,” he remarked. Copilot can also alert the user if outside information is needed.

Ultimately, case studies from the conference demonstrated how AI technology has the potential to make engineers more productive and ensure the safety and reliability of infrastructure. As Anne-Marie Friel, a Partner at Pinsent Masons, emphasized, particularly in a high-risk sector like water, decisions must be explained and verified. As Cumins said, “AI and infrastructure will remain a collaborative process with a human end.”

Christian Bonawandt is an industrial content writer for Water Online. He has been writing about B2B technology and industrial processes for 24 years.