Water utilities around the world are at a critical juncture. Climate change, urban growth and aging infrastructure are putting unprecedented strain on already limited and precious resources, of which, water is key. Research from the World Resources Institute’s Aqueduct Water Risk Atlas puts at least 50% of the world’s population, about 4 billion people, at high risk of facing water stress each year1. These risks are not theoretical or in the distant future - they are evidenced today through failing pipelines, service disruptions and escalating operational and maintenance costs.
This blog is part of our 'AI for Water' series, where we explore how artificial intelligence can transform water utilities. In this perspective, we delve into the practical applications of predictive AI and how it can address some of the water sector’s most pressing challenges, from aging infrastructure to operational inefficiencies.
The State of Play and Role of Predictive vs Generative AI
In Europe, systems built over a century ago are grappling with meeting modern infrastructure demands. In the United States, the American Society of Civil Engineers (ASCE) has consistently graded the nation's drinking water infrastructure at a "C-" level, indicating significant deficiencies. Further, in India, water demand is projected to double by 2030 due to population growth.
Despite these pressures, many water utilities still continue to rely on outdated systems that are costly, inefficient and vulnerable. Failure to innovate could mean widespread disruptions, and moreover, result in huge economic and social costs.
-
READ MORE
Artificial Intelligence (AI) offers a way forward. If applied effectively, AI could revolutionize how water utilities manage limited resources and optimize operations. While much of the conversation today evolves around generative AI - a technology that creates content and has sparked debates around ethics and risks - predictive AI is a more promising solution, operating quietly in the background, delivering tangible results.
Unlike generative AI, predictive AI has been around for decades and has matured significantly over the past 10-15 years. Its purpose is clear: to forecast, optimize and improve systems based on data-driven insights. What’s more, the risks associated with predictive AI are far less, making it a more practical and reliable solution for water utilities. And, it focuses on critical applications needed today like pipe condition assessment, forecasting future system states and optimizing resource utilization.
With the urgency to act, now is the time to move from theory to practical application.
This is not the future; it’s happening now. We’re past the point where AI is a theoretical solution. It’s a necessity. AI offers water utilities the ability to predict problems before they occur, optimize the use of limited resources, and most importantly, ensure service reliability even in the face of climate change and regulatory demands.
A Lifeline for the Aging Infrastructure Crisis
Aging infrastructure remains one of the most significant challenges for water utilities worldwide. A critical consequence is non-revenue water – treated water that is distributed but lost before it reaches customers due to leaks, theft or billing errors. This not only results in billions in lost revenue globally but also leads to waste of valuable energy and resources used to pump, treat, and distribute water, directly contributing to avoidable carbon emissions. The financial consequences add to billions in revenue lost across the globe. In Europe, an umbrella group of water suppliers, EurEau estimates that about 25% of all water distributed is lost, with nearly all of this attributed to leaky pipes2. While in the UK, this has led to a staggering 3 billion liters of water lost every day3. And in the US, ASCE's 2021 Report Card highlighted an estimated 6 billion gallons of treated water lost daily through aging and deteriorating pipes, costing utilities approximately $2.6 billion annually.
These losses are not only a financial burden but also a critical resource challenge, as utilities struggle to meet increasing demand amid climate pressures and growing urban populations. The sheer scale of water loss warrants urgent action to address aging infrastructure. Yet, traditional methods of repair and maintenance are often costly and too slow to implement. Furthermore, taking a reactive approach fails to prevent systemic issues or anticipate vulnerabilities. Instead, utilities must embrace smarter, more proactive strategies that tackle the root causes of water loss.
Predicting failures before they happen
Aging infrastructure is an issue that will only get worse unless we get smarter about how we manage it. Predictive AI enables utilities to quickly identify potential weak points in their networks, prioritize maintenance efforts and prevent failures before they occur. By leveraging real-time monitoring and advanced analytics, utilities can minimize water loss and secure their financial and environmental future.
In the UK, the High Speed 2 (HS2) project requires significant water utility and management efforts, including major asset diversions or rerouting,, flood protection and river management. These were supported by advanced analytics through Arcadis’ EDA platform, leveraging artificial intelligence (AI) and machine learning (ML) to ensure data-driven decisions in planning and execution.
By integrating AI and ML through our EDA platform, we are enabling HS2 to tackle complex water management challenges with precision and foresight. This technology analyzes vast datasets, identifying planning patterns and simulating uncertainties across various scenarios. In doing so, it allows us to prioritize actions that mitigate risks, optimize resource use and enhance resilience for both immediate and future challenges. It’s a testament of how digital innovation can drive smarter, more sustainable infrastructure solutions.
In the US, Arcadis partnered with the San Antonio Water System to implement an AI-driven asset management system that analyzed variables such as pipe material, age and environmental conditions to predict failure risks. This proactive approach is helping the utility prevent costly disruptions and reduce water loss, enabling them to prioritize investments where they are most needed and extend the lifespan of critical infrastructure.
Accelerating intelligent and responsible infrastructure design
Intelligent water solutions can also support smarter infrastructure planning. Generative design tools allow utilities to simulate multiple design scenarios, incorporating resilience and efficiency into every option. These tools such as Arcadis’ Asset Generator, which is a cloud-based tool to automate the design and creation of 3D models and facilitates master planning in a GIS environment, speed up the planning and design process while enabling utilities to future-proof their investments, ensuring long-term viability.
Transforming Water Management and Service Delivery
Predictive AI is not just about leveraging advanced technology because of the ‘hype’ - it’s about accurately pinpointing and addressing long-standing operational inefficiencies and environmental challenges in water utilities. If used right, it can help utilities transform scattered data into actionable insights that drive efficiency, optimize resources and minimize risks.
Operational efficiency is the low-hanging fruit where AI can deliver immediate results. From reducing energy consumption to optimizing how we use chemicals in treatment, AI enables us to do more with less. This translates directly to savings that can be reinvested into the system.
So, who stands to benefit? Water Treatment and distribution system operators can leverage AI to better manage water age and energy use across their networks. Collection system and wastewater treatment plant operators can make informed operational decisions translating weather observations with AI into highly accurate flow forecasts. These forecasts allow assets to be brought online proactively and controls to be set at optimal points, maximizing treatment efficiency and minimizing spills.
Process and quality optimization beyond reducing outages
Reducing outages is one of many benefits of predictive AI, but the true potential lies in its ability to optimize operations and quality across the entire utility.
The power of AI lies in its ability to make sense of vast amounts of data in ways humans simply can’t. By leveraging AI, we’re allowing utilities to move from reactive to proactive modes of operation - saving time, money and critical resources, while improving overall system quality.
AI empowers utilities to:
Anticipate weather impacts:
Predict stormwater surges and adjust controls to prevent overflows or treatment disruptions.
Enhance water flows:
Forecast demand and optimize pumping schedules to ensure consistent supply and minimize energy usage.
Improve water quality:
Detect and mitigate potential quality issues before they escalate, ensuring compliance and safeguarding public health.
In Southern California, Arcadis implemented an AI-powered predictive modelling system that integrates machine learning with an advanced multispecies water quality model (built in Autodesk’s InfoWater Pro platform) to monitor disinfection byproducts (DBPs). Predicting potential spikes enabled the utility to adjust its processes in real time, achieving compliance while delivering better water quality to customers.
Proactively managing greenhouse gas emissions
Reducing greenhouse gas emissions is a pressing challenge for utilities today, particularly nitrous oxide (N₂O), which is 300 times more harmful than carbon dioxide (CO₂) and also contributes to ozone layer depletion. Unlike CO₂, N₂O emissions cannot simply be offset. This requires them to be addressed directly at the source, making their accurate prediction and management critical for utilities striving to meet climate goals.
Arcadis partnered with Welsh Water to implement an AI-powered system designed to accurately predict N₂O emissions. By collecting process data from six wastewater treatment plants (WWTPs) and feeding it into Cobalt Water’s AI (a machine learning-based platform), the system was able to to forecast N₂O emissions using operational data alone. The predictions were then corroborated with test hoods using N₂O sensors. The result was a phenomenal correlation of up to 90% accuracy between the AI model and physical measurements.
This breakthrough represents a significant advancement in wastewater treatment. Using only operational data, the AI system was able to predict and proactively manage one of the most harmful greenhouse gases. In the pre-AI era, such precision was unattainable. Now, utilities can take targeted actions to minimize emissions as part of their process optimization strategies, aligning with regulatory requirements, while delivering far better water quality for customers.
Overcoming Regulatory Complexities, Enhancing Compliance
As utilities embrace AI, the regulatory environment presents both challenges and opportunities. Governments and regulatory bodies across all regions of the world are increasingly scrutinizing AI to ensure ethical deployment, operational transparency and data security.
The European Union’s Artificial Intelligence Act, one of the most comprehensive regulations for artificial intelligence that came into effect from 1 August 2024, mandates stringent requirements for transparency and accountability. For water utilities, this means ensuring that AI models used for predictive maintenance or water quality forecasting are safe, transparent, traceable, non-discriminatory, environmentally friendly and overseen by humans. Similarly, while AI legislations across the U.S. is at various stages, over 30 states have some form of proposed and/or enacted legislation leading to strengthened cybersecurity measures for critical infrastructure to safeguard against data breaches and malicious attacks.
Integrating AI into water utilities often involves navigating these regulatory complexities. Arcadis conducted independent research in coordination with state regulatory agencies across the United States to understand their stance on AI adoption. Responses varied widely - some indicated that AI use is not explicitly addressed, while others advised that its implementation will require an engineering plan review due to the potential impact on treatment plant effluent quality. These regulatory nuances emphasize the importance of balancing innovation, responsible deployment and compliance.
Turning Compliance into a Competitive Advantage
Predictive AI can help utilities not just meet regulations but often exceed them.
The regulatory environment is constantly shifting, and AI gives utilities the agility to keep up. Whether it’s identifying lead service lines or optimizing treatment processes, AI helps us ensure compliance in a world where the goalposts are always moving.
While regulatory requirements may often seem burdensome, they also offer an opportunity for utilities to lead the way in ethical and transparent AI deployment. Predictive AI can act as a compliance enabler. For example:
Data-Driven Reporting:
AI simplifies the process of generating compliance reports by integrating real-time operational data with regulatory requirements.
Risk Management:
Predictive models can identify and mitigate risks before they lead to regulatory violations.
Enhanced Accountability:
Transparent AI systems demonstrate a utility’s commitment to ethical operations, building trust with stakeholders.
Preparing the Workforce for an AI-Driven Future
While AI holds great promise, it also presents a challenge: workforce readiness. Utilities are faced with a two-fold problem. On one hand, older workers are retiring at an increasing rate, leaving utilities struggling to fill critical positions. On the other, adoption AI requires a workforce that understands its capabilities and knows how to use it to maximize its potential.
The adoption of predictive AI is as much about people as it is about technology. Without the right workforce strategies, even the most advanced AI systems cannot deliver their full potential. Similar to how the pandemic accelerated the adoption of remote work, these pressures may normalize the use of AI to perform tasks that were once considered too risky or unconventional.
Investing in upskilling
Arcadis is working with utilities in Germany to develop targeted AI training programs. These initiatives ensure workers at all levels are upskilled and equipped with the right skills needed to integrate AI into their daily operations.
AI is not going to replace workers, but it will change the nature of their work. Many in the water workforce will be able to work more efficiently using AI to support decision-making. We need to ensure that the current workforce is trained to take advantage of these new tools. Upskilling is not just an option - it’s a necessity if we want to realize AI’s full potential.
Empowering a new workforce dynamic
As older systems and work practices give way to AI-driven operations, utilities need a new breed of workers: who are not only familiar with water systems but also comfortable with data analytics and machine learning.
Here, we explore the top three fastest growing occupations we predict and evolution of traditional roles the increasing adoption of AI:
Data Scientists and Information Security Analysts:
Specialists ensuring seamless data integration, cyber security and algorithm refinement.
Mid-Level Technicians & Systems Managers:
AI becomes a decision-support tool, helping technicians prioritize maintenance and respond proactively to anomalies.
Senior Engineers and Managers:
Leaders use AI to inform capital planning, regulatory compliance and strategic decision-making.
Our next blog will explore this transformation in more detail, offering actionable strategies for building a future-ready workforce.
From Hype to Real Solutions: The AI Blueprint
It’s easy to get swept up in the hype surrounding AI, but the real power of this technology lies in its ability to solve specific, pressing problems for utilities.
Many have jumped into AI initiatives without a clear plan, only to find that the technology fails to deliver on its promises. But when applied thoughtfully and strategically, AI can provide real, measurable results.
The integration of predictive AI is a journey, not a destination. So what does the AI Blueprint to enhance process efficiency in Water Utilities look like?
Here’s a step-by-step guide when considering AI for your utility:
-
Step 1: Assess Your Digital Maturity
Before implementing AI, evaluate your digital ecosystem to establish a strong foundation. It’s not enough to have multiple data points. What matters is that they are ‘relevant’ data points that connect up with your needs and line up in the right ecosystem. Consider:
- Data accessibility: Is your data siloed, or can it provide a unified view of operations?
- Data quality: Is the data reliable, clean relevant and accessible for predictive analysis?
- System readiness: Are existing systems and processes compatible with AI integration?
Developing a roadmap to fill gaps in data collection and digital infrastructure will be critical for AI success.
Arcadis offers Digital Maturity Assessments, helping utilities evaluate their current capabilities and pinpoint areas for improvement. This structured approach provides a roadmap to prepare systems and processes for AI adoption.
-
Step 2: Define ‘High Impact’ Use Cases
AI can solve a range of issues, but prioritization is key. To ensure a focused and effective rollout, utilities should start with high-priority challenges that offer the greatest, measurable benefits. Consider:
- Pressure areas: What are the most critical operational inefficiencies (e.g., leak detection, asset failure, energy use)?
- Regulatory considerations: Are there specific regulatory pressures or environmental targets that AI could help address?
- The benefits case: Identify high-impact areas where AI can improve service reliability, reduce downtime, or optimize energy and resource use.
Here are a few examples of ‘high impact’ use cases:
Water Loss Reduction:
Identifying and addressing leaks to minimize non-revenue water.
Pipeline Failure Prediction:
Using AI to forecast risks and prioritize preventive maintenance.
Energy and Chemical Optimization:
Reducing operational costs by enhancing efficiency in treatment processes. -
Step 3: Select the Right Tools and Partners
The market is flooded with AI tools, making the right choice difficult. The success of AI implementation depends on choosing tools and partners that align with operational needs. Consider:
- Scalability: Can the tools expand to meet future demands?
- Compatibility: Do the solutions integrate with your existing systems and operational needs?
- Expertise: Are there experienced implementation partners (like Arcadis) who can guide you on tool selection and water-specific AI applications that fit your context?
Arcadis offers a comprehensive suite of intelligent water solutions designed to address these considerations. Our approach focuses on transforming water utilities into fit-for-future entities by leveraging digital technologies to improve financial stability, customer experience and operational performance.
-
Step 4: Pilot and Scale Strategically
Pilot programs are essential to test AI systems, measure their impact on performance and refine strategies ahead of full-scale deployment. Consider:
- Establishing clear KPIs: Track metrics such as cost savings, efficiency improvements or service reliability.
- Start small: Focus on high-risk or high-impact areas for initial implementation.
- Iterate and improve: Use pilot results to refine the approach and build confidence before scaling.
- Workforce preparedness: Ensure your teams have the right digital skills and literacy to adopt the new tools
Arcadis’ Water Finder creates a real-time digital representation of utility networks, enabling visualization, forecasting and optimization of operations, helping utilities test scenarios and make informed decisions. What’s more, utilities can participate in a ‘pilot testing program’ to establish if AI is the right fit.
-
Step 5: Monitor, Optimize and Improve
Predictive AI is not static - it improves over time with new data, regulations and operational needs. Utilities must establish a feedback loop to continuously monitor system performance and refine AI models, ensuring they remain adaptable to emerging challenges. Consider:
- Real-time performance tracking: Use analytics to monitor system efficiency and identify bottlenecks.
- Model refinement: Regularly update models with new data and established feedback loops to ensure accuracy and continuous improvement.
- Regulatory alignment: Ensure AI systems remain compliant with evolving regulations.
Unlocking the Future of Your Utility with AI
Predictive AI is already transforming the water industry, offering practical solutions to urgent challenges. And while it’s a powerful tool, it’s not a standalone solution and needs to be part of a broader strategy. With the right infrastructure, tools and workforce in place, utilities can move beyond reactive problem-solving to proactive management, building smarter, more sustainable systems for the future. Now is the time to start building a roadmap for AI integration.
As your sustainable transformation partner, Arcadis combines decades of expertise with cutting-edge technology to guide you through this journey. Whether you’re starting from scratch or looking to optimize your current systems, our intelligent water solutions are tailored to address your unique challenges - helping you deliver operational efficiency, ensure compliance and secure the trust of your stakeholders.
Ready to reimagine what’s possible for your utility? Connect with an Arcadis expert today to explore how we can help you adopt AI to not just solve today’s challenges but lead the way in shaping the future of water management.
Stay tuned for our next perspective in our ‘AI for Water’ series, where we explore in detail workforce readiness and ethical use of AI in the water sector.
AI for Water: A game-changer or a pipe dream?
Paired with human intelligence, AI has the power to tackle many of the water sector’s most pressing problems. But, what are the key building blocks to be considered when it comes to AI and water?
