Water utilities know where AI fits — but can the foundations carry it?

AI dominated the floor at Ozwater 2026, but many of the conversations kept returning to a more basic question: whether the digital foundations beneath it — connected systems, trusted data, governance — are ready to carry it.

Over three days of conversations at Ozwater 2026 with operators, engineers, analysts and asset managers, the same theme kept surfacing: real enthusiasm for AI, paired with real uncertainty about whether the underlying systems are ready for it.

Australian water utilities have been collecting data for decades. What many lack is the foundation to use it: connected systems, trusted data, governance to act on it with confidence.

The ambition within the sector is genuine. Utilities described wanting to utilise AI for earlier leak detection, more accurate demand forecasting, better asset performance and treatment optimisation. Yet the constraint attendees kept describing had little to do with the models themselves. It came back to the information feeding them, and whether it was current and reliable enough to support an operational decision.

A consistent gap between appetite and readiness

The pattern held across utilities of varying sizes: a strong interest in AI that rested on data foundations that are not yet ready to carry it. This was widespread rather than confined to any one type of organisation, and it ran deeper than a handful of isolated data quality fixes.

In one organisation, performance reporting ran on an eight-day lag, assembled by hand across several delivery partners and dozens of separate reports. Elsewhere, early AI pilots had stalled. The technology was rarely the issue. The data beneath it was unclassified, poorly governed and scattered across systems that did not talk to each other. Engineers talked about hunting for documents they could not find; analysts talked about results they were not willing to put their name to.

For most utilities, the difficulty wasn’t the AI use cases themselves. Many had already identified where they wanted to apply AI, and some were building prototypes themselves. The hard question was whether they had the digital foundations to support those ideas safely and at enterprise scale. Without those foundations, AI tends to expose and amplify the gaps already there rather than resolve them.

None of this is abstract. It plays out in daily operations: field crews with little visibility of work already under way, customer results going out with missing or incorrect details, and teams patching the gaps between systems with spreadsheets, email and manual re-keying. Layer AI over an environment like that and it tends to magnify the existing risk rather than remove it.

Where the most valuable data gets stuck

Control room and SCADA teams will recognise the pattern. Some of the most valuable operational data sits in specialised systems built for exactly that purpose, and that specialisation is a strength rather than something to engineer away.

The problem is that this data often doesn’t flow cleanly into forecasting, reporting or customer systems. The systems can stay separate and specialised; it’s the data between them that needs to interconnect. Today, what bridges them is usually a manual layer of re-keying and duplication, and what comes with this is delay. That delay matters most in the moments operators are already under pressure. For example, managing a flood event, triaging several contamination alerts at once, or weighting maintenance and efficiency against environmental impact.

AI is only ever as useful as the environment it runs in. While the underlying systems stay disconnected, ‘real time’ stays aspirational and the data remains hard to trust. In water operations, where a decision can carry immediate consequences, that gap is not a side issue for the IT team to tidy up later. It shapes whether AI can be used at all.

Regulation raises the bar

This carries extra weight because water and sewerage operators are designated critical infrastructure under the Security of Critical Infrastructure Act 2018, which sets obligations around risk management, cyber incident reporting and the protection of essential services. In that setting, decisions about AI are wrapped up tightly with governance, accountability and public trust as much as with the technology itself.

It also explains why the ‘move fast’ instinct from other sectors sits awkwardly in water. Utilities carry obligations across public health, billing integrity, regulatory reporting and service continuity. Automation that can’t be explained or audited is unlikely to win the confidence of a board or a regulator, let alone the public.

Doing the work in the right order

For many utilities, the more useful next step is not a further AI pilot but the slower work of integration. This means connecting operational, customer and enterprise systems so information can move between them cleanly and in context, without people translating it by hand. It rarely makes for a compelling demo, but it is what allows anything built on top to scale.

The utilities that get the most from AI tend not to be the ones running the boldest pilots. They are the ones that sequence the work by establishing trusted foundations, governing the data, proving value in a controlled setting, and only then extending automation into higher-stakes decisions.

There is nothing anti-innovation about working this way. Sequencing is what gives innovation something solid to stand on. In a sector where accountability can’t be handed to an algorithm, getting the order right is very much about having a sound AI strategy.

From data-rich to decision-ready

The water utilities sector has already put real money into capturing data. The work now is making that data usable and reliable enough to support better decision-making. Interest in AI is no longer in doubt; what will set utilities apart is their willingness to do the less visible work that using AI responsibly demands.

For a sector under pressure to modernise without putting essential services at risk, that willingness may matter more than any single technology choice. The utilities organisations that invest in trustworthy foundations first will be the ones able to get real value from AI when they turn to it.

Image credit: iStock.com/Yuuji