Infrastructure AI comes of age in the global water sector

A global survey released during the event, carried out with Pinsent Masons, Mott MacDonald and Turner and Townsend, suggested that about half of infrastructure professionals are already piloting or implementing AI. Around one third of submissions to this year's Going Digital Awards and nearly half of the finalists reported some use of AI. Against that backdrop, Bentley Systems used the conference to set out its infrastructure AI strategy and invite users into a wider collaboration.

Infrastructure AI and the role of context

For the water sector, the idea of infrastructure AI is closely linked to existing modelling tools. In an interview on the sidelines of the event, Francois Valois, Senior Vice President, Bentley Open Applications at Bentley Systems, explained that Bentley is positioning infrastructure AI as something built on its established analytical engines.

A global survey released during the event suggested that about half of infrastructure professionals are already piloting or implementing AI

He noted that at this year's Conference, the company announced the Infrastructure AI initiative, focused on co-creating solutions with customers by leveraging existing Bentley tools and application programming interfaces. As an example, he pointed to the OpenFlows engine, widely used for analysing water networks during planning and design. Although its APIs were not originally created for AI, they can now act as the numerical engine for AI-powered capabilities. Until now, he added, some of these possibilities have been constrained more by business models and licensing terms than by technology. The intention of the initiative is to work with users rather than discourage experimentation and to support innovation across all of Bentley's industry segments, with particular relevance for water, where OpenFlows is widely used.

Molly Brown, Senior Director of Product Management at Bentley Systems, stressed that in water engineering, precision is non-negotiable and mistakes have real consequences, which is why humans must remain in the loop and AI models need to be trained using project, historical, environmental and engineering context. In her view, AI should help eliminate mundane, repetitive work rather than the workforce, but human oversight remains essential because of the complexity of water systems.

Digital twins, WaterSight and managing data volume

Turning context into operational insight is a central challenge for utilities. In the water sphere, this often centres on digital twins and on integrating hydraulic models, supervisory control and data acquisition systems and other monitoring tools.

In water engineering, precision is non-negotiable, and mistakes have real consequences, which is why humans must remain in the loop

Valois pointed to Bentley Systems' work with OpenFlows WaterSight as an example of how this is being approached. He described WaterSight as a tool that supports operational water utility management and has been deployed in Brazil and other regions. It is AI-infused, including alarm and anomaly management across SCADA and IoT data sources. In practical terms, this means that when water pressure suddenly drops and a leak alarm is triggered, AI can analyse historical context and help distinguish genuine leaks from false positives linked to other events such as heat waves. Given the size and noise of these datasets, Valois argued that it is unrealistic for humans to interpret them manually and that AI can filter noise and point operators to specific locations worth investigating.

He also acknowledged that creating and maintaining a digital twin is challenging in its own right. AI can assist by improving input data quality and automating repetitive tasks, such as schema mapping between incoming data formats and internal systems, opening up many possible applications.

In a separate interview, Cecilia Correia, Global Water Industry Solutions Strategist at Bentley Systems, underlined the role of hydraulic modelling as a starting point for this journey. She said that hydraulic modelling, which sits at the core of OpenFlows, is essentially an early form of a digital twin, a digital representation of the physical network. According to her, it allows engineers to understand how their systems behave under different or critical conditions, which is essential for resilience and sustainability, for example, when preparing for future challenges such as population growth or increased water demand.

Looking ahead, she identified real-time modelling as the next step, by connecting the digital twin to live data from pumps, treatment plants and sensors. In her words, that transforms the model from something static into a dynamic system that helps utilities detect issues, optimise operations and plan for future resilience. She described this as an infinite loop rather than a closed circle, a continuous process of learning, adapting and improving.

Openness, interoperability and data ownership

If digital twins are to work across multi-stakeholder projects, the question of openness is central. In a joint interview, Oliver Conze, Senior Vice President and Head of Product for Bentley Infrastructure Cloud, and Jason Slocum, Director of Product Management at Bentley Systems, emphasised that open standards and data access policies have become key parts of the company's positioning.

Slocum described interoperability as a permanent differentiator. He observed that across design, construction and operations organisations are seldom working with software from a single vendor, nor with a single firm. In that context, he argued that an open ecosystem capable of bringing information together, aligning and aggregating it and providing a single view of truth is essential for activities such as design reviews and 4D construction planning.

Furthermore, Slocum highlighted that openness is a core pillar of what Bentley does, covering both the ability to ingest data and to export it using open standards. He stressed that the company strongly believes that customer data belongs to the customer, so it does not lock information into Bentley or Bentley Infrastructure Cloud. Data remains available through open standards such as IFC.

Those points were echoed in the keynote, where Cumins reiterated that users' data remains their own and explained that the company trains AI models only on licensed or voluntarily contributed data, with a Data Agreement Registry intended to provide visibility into how information is used.

For utilities and their regulators, the question of trust goes beyond licensing detail. Valois linked it directly to professional responsibility. He commented that AI can help engineers and operators become more productive but cannot replace them today and that faster processing enables optioneering by giving engineers more time to explore alternative solutions and consider factors such as environment, resilience, compliance, security and safety. He argued that engineers should embrace AI, learn its strengths and limitations and apply it responsibly rather than relying on it for every task.

Slocum also pointed out concerns about data trust from a product perspective. He said that while there is excitement around AI, there is also hesitation, as users want clarity on how their data is used. He explained that Bentley emphasises that data remains owned by the customer, is not used to train models without permission and that AI usage is auditable, arguing that trust is critical because organisations want to innovate but must protect sensitive information.

Adoption, culture and the platform question

Across the various interviews, the question of adoption surfaced repeatedly. When asked about barriers to scaling smart water solutions, both Correia and Richard Vestner, Vice President at Bentley Systems, placed more weight on culture than on technology or finance.

Creating and maintaining a digital twin is challenging; AI can assist by improving input data quality and automating repetitive tasks

Vestner described adoption as the key challenge, noting that technology exists and is scalable, and that usability is addressed through user experience and design. In his view, digital approaches also require trust, not necessarily in a fully autonomous machine but in tools that remain under human control. He explained that Bentley is trying to make technology transparent rather than a black box, but that users still need curiosity and willingness to try new methods. He characterised technology as something that should augment what people do, offering more options and inspiration, and suggested that digital twins in particular can support faster and better decision-making by making it easier to test what-if scenarios.

Correia added that fears of job loss remain an obstacle even where utilities struggle to recruit. She observed that many people worry about being replaced by machines, yet when speaking with utilities, the first thing they mention is a shortage of staff. In her view, roles will change rather than disappear, with a shift towards profiles such as data scientists and away from manual tasks such as physically operating gates. She argued that more people will focus on validating machine outputs, running tests and understanding accuracy and that this represents upskilling rather than replacement.

He said he hoped that products and platforms would drive more standardisation and integration so that convergence between sectors would become possible. As an example, he referred to Bentley’s concept that any product can share and leverage data in one place what Bentley calls an iTwin, a cloud-based container of information. The more this information is brought together and kept up to date, he argued, the more domains, specialists and sectors can participate and learn from each other, increasing transparency.

Co-pilots, co-innovation and the next steps for water

Throughout the event, Bentley presented a growing list of AI-enabled applications, including OpenSite+, OpenUtilities Substation+, and SYNCHRO+, all of which are connected to Bentley Infrastructure Cloud. Valois described these as early examples of how generative tools are being embedded in engineering workflows.

AI enables smaller firms to compete with much larger organisations by allowing them to move more quickly with a start-up-like mentality

He recalled that Bentley was the first to use a large language model in engineering through OpenSite+, which the company had showcased at the previous year's conference. Initially, he said, it was seen as a potentially useful but limited tool, but it became more impactful than expected. This experience led to the creation of Bentley Copilot, a shared AI component that can be applied across applications. Valois stressed that context remains important and that each version is specialised using a technique known as Retrieval Augmented Generation so that the model remains within the appropriate technical scope. For Slocum, the main value of these co-pilots lies in their ability to remove routine workload. Speaking from his experience as a civil engineer, he referred to the amount of repetitive work involved in labelling, searching for files and managing documentation and argued that co-pilots can help users be more productive and do more with less, particularly in the context of workforce shortages. He cited a remark he had heard that AI will not replace people, but that people who use AI will replace people who do not and suggested that there is some truth in that observation. In his view, AI can also enable smaller firms to compete with much larger organisations by allowing them to move more quickly with a start-up-like mentality, while co-pilots help remove tedious tasks so that engineers can focus on value-adding work.

For the water sector, the interviews and discussions in Amsterdam suggest that the next phase will combine more advanced digital twins, AI-driven analytics, and strengthened communication and governance. As Correia noted, emerging technologies like AI will have a major impact, but the sector is still in the early stages of understanding their full potential. She emphasised that while the opportunities are significant, technology alone is not enough. Real progress will depend on people who understand both the critical value of water and the responsibility of managing it, and who can help drive the cultural change needed for adoption.