Artificial intelligence (AI) is already transforming the water sector, enhancing efficiency, reducing failures and optimising operations. But as adoption grows, so do questions about sustainability, ethics and long-term impact.
To celebrate National Science Week, we’re sharing a recent Q+A with SpiralData Chief Data Scientist Dr Ram Balachandran on AI’s role in the water sector. Balachandran says that while AI’s role in water management is still evolving, its potential to improve resilience and conserve resources is undeniable.
When we talk about AI, people often think of large language models like ChatGPT, but AI has been around for a long time. Essentially, AI is about pattern matching – identifying patterns in data, validating them and then productionising them to extract value.
In asset-heavy sectors like water, AI improves efficiency in multiple ways – labour efficiency, energy efficiency and water conservation. There’s also the renewal of capital investments, including finding ways of prolonging the life of an asset, ensuring ongoing operations and minimising impacts.
At SpiralData, we focus on transient detection. Transients are pressure surges in pipes; they are a major problem, responsible for about 50-60% of asset failures. If we can reduce unexpected failures by even 5%, that significantly shifts the economic outcome. We detect pressure surges, identify the root causes, measure their impact on pipes, and work to mitigate them to reduce leaks and breaks. This also means fewer emergency callouts for repairs, which is a major operational benefit.
Other companies are also using AI for efficiency gains, including applying AI for asset inspections. Traditionally, engineers would spend hours reviewing footage to identify cracks and leaks, but AI-driven computer vision can automate much of that process.
Energy efficiency is another area where AI has potential. In wastewater treatment, blowers consume significant energy. AI can help optimise energy use by responding to demand, water quality and other factors. AI is making its way into the sector, but unlike a simple chat interface, its adoption is still at a technical level. Over time, solutions will become more holistic, helping both capital efficiency and operational efficiency.
Even for end consumers, AI could help prevent issues before they happen, making water management more predictive rather than reactive.
AI can help predict asset usage more effectively. Many network assets – such as 100-year-old pipes and valves – are not fully utilised because their state is unknown.
There is significant opportunity for AI-driven network optimisation. Traditional numerical simulations struggle to run complex scenarios, but AI can model these efficiently, helping to optimise how pumps and valves interact.
Currently, most AI work in water focuses on individual assets, but there’s still untapped potential at the network level. By creating digital twins – virtual models of water networks – we can optimise systems holistically rather than in isolated parts. This shift to systemwide efficiency is what I find most exciting about this work.
I think the key question is: how much energy or water does AI consume versus how much does it save?
The AI used at SpiralData is not energy intensive. We run it on CPUs (central processing units), rarely needing GPUs (graphic processing units) except for initial training. Most of the processing runs on low energy cycles.
More importantly, the energy savings from extending asset life, preventing leaks and improving operational efficiency far outweigh the AI’s energy consumption. For instance, undetected leaks lead to significant water wastage and unnecessary pumping costs. AI helps mitigate that, reducing both water and energy waste.
Regarding ChatGPT and large AI models: yes, they have high energy demands. There is also a considerable amount of water required for cooling in data centres.
However, improvements are underway – miniature AI models can now run on laptops with far lower power consumption. AI usage should be judged case by case: what problems does it solve and is it worth the trade-off?
For the water sector, AI provides tangible savings in water, energy and emissions, making it a net positive rather than a drain on resources.
This is something I think about a lot. There are two major concerns. The first is human oversight. AI does not inherently understand the laws of physics or conservation.
That’s why AI should always be a decision-support tool, not an autonomous decision-maker. The responsibility for final decisions must always rest with a human expert. AI can provide insights, but it should not be making operational control decisions, such as adjusting valves or pumps.
The second is data privacy. Water-use data is incredibly sensitive. It can reveal personal behaviours, and it must be carefully anonymised and aggregated before being used in AI models. Data governance frameworks in Australia and Europe are improving, but they must also evolve to address AI-specific concerns.
In my opinion, the water sector must ensure strong data governance while keeping humans in control of AI-assisted decision-making.
AI is a powerful forecasting tool. For climate change, AI can model various future scenarios, predicting which areas will be most affected. Sydney Water, for instance, is dealing with salt accumulation in sewer outlets due to salt-water ingress. AI can help forecast such impacts and develop mitigation strategies.
For population growth, AI can simulate how increased demand will stress existing water networks, helping planners optimise infrastructure investment.
There are also advances in AI-driven weather forecasting. Google has developed WeatherNext, which forecasts weather as accurately as traditional numerical simulations, but at a fraction of the computational cost. These improvements will extend to climate forecasting and water planning.
The major challenge isn’t the technology – it’s human collaboration. Different stakeholders need to align on worst-case scenarios and mitigation strategies. The best AI tools are useless if people don’t understand or trust them.
This is why AI literacy across the water sector is essential. AI is just another tool, like engineering models or numerical simulations. The real challenge is getting people to use it effectively.
I’d flip the question around: what is the cost of not adopting AI? Water is a conservative sector, for good reason – safety is paramount. But there is also a risk in not adopting new tools. AI will never be perfect, but waiting for it to be 95% accurate (like traditional engineering tools) may mean missing out on valuable efficiencies.
The water sector should focus on education and improving AI literacy so that professionals understand its capabilities and limitations. For example, TAFE NSW has created courses to introduce AI to people from all backgrounds.
But incremental adoption is also an important focus. Running small AI trials across different use cases to build confidence and refine best practices is a safe way to work through new approaches. Iterative improvement is another approach that should be kept front of mind in the years to come.
AI implementation should be an ongoing process, continuously refining approaches rather than waiting for a ‘perfect’ solution. The goal isn’t to let AI take over but, to integrate it responsibly, making small, iterative improvements that lead to major efficiency gains over time.