The infrastructure sector is not known for moving quickly. Bridges, tunnels and rail lines are designed to last a century or more, with the engineers who build them are trained to be methodical and precise. Nevertheless, AI is beginning to change the pace – and the nature – of the work. 

Julien Moutte, Chief Technology Officer at Bentley Systems, which develops engineering and infrastructure design tools used on projects worldwide, is at the forefront of this transformation. He argues that AI presents an opportunity to do more meaningful work.

“There is a massive demand for infrastructure and it’s not going to slow down,” Julien contends. “Climate change, population growth and conflict around the world mean there is not going to be a shortage of work anytime soon. What we know for sure is that we see a shortage of civil engineers.”

That shortage, Julien believes, makes the case for AI as a force multiplier – enabling fewer engineers to achieve more, faster and to a higher standard.

Collaboration at scale

Bentley Systems works with major engineering firms rather than directly with infrastructure operators. Its software underpins some of the most complex construction projects in the world, from airport expansions to urban rail networks.

One prominent example is Crossrail – the vast tunnelling and rail project in London that opened as the Elizabeth line in 2022. The project involved dozens of engineering firms working simultaneously on different sections of the route.

A project of that complexity requires a shared digital environment where models, drawings and data from multiple contractors can be reviewed together. 

“The owner and operator will want to procure the data from the different companies engaged on a project,” Julien explains. “They need a place where they can bring all of the 3D models together, review them in a coherent and consistent way, and coordinate the different people working on it.”

Without such a system, engineers would be exchanging paper drawings and trading digital models in unwieldy formats. The use of a common data environment – a single platform where all project information is held and accessed – made collaboration at that scale possible.

AI can now extend what such platforms are capable of. Rather than relying on human reviewers to check whether designs from different firms are consistent and meet quality standards, an AI agent can run those assessments automatically.

“You can have an AI agent connect into that common data environment and run quality validations at a much faster pace than humans,” Julien goes on, “allowing teams to detect when something is not up to the expected standard and remediate this before it becomes a bigger problem.”

Smarter scheduling at Heathrow

At Heathrow Airport, major construction work must happen around a fully operational facility, minimising disruption to passengers. 

Bentley’s Synchro software, used for construction scheduling and planning, has been deployed on expansion and refurbishment projects there. The tool allows engineers to attach detailed schedules to a 3D model of the construction work, so that every task is planned in sequence. 

The latest version, Synchro Plus, brings AI into that process. The software can automatically assign tasks to a logical order and recalculate the entire programme in real time when circumstances change. 

Julian adds: “When something changes, AI is able to recompute the schedule taking into account parameters like the weather forecast, which has a big impact on pouring concrete.”

Giving AI an engineering licence

The phrase Julien uses for Bentley’s approach to AI-generated design is striking: “giving an engineering licence to AI”.  

What he means here is that AI outputs should be validated by the same proven engineering tools that human professionals rely on.

“When an engineer needs to validate their work, they run a structural analysis with tools that have been proven in the market as reliable,” he explains. “Our approach is to team up those tools with AI agents, which allows us to give an engineering licence to AI so that it can help engineers by running alternatives and validations of a design.”

This matters because engineers are professionally accountable for the designs they sign off. 

“Nobody wants to sign their name on a design if they’ve not been able to verify what was generated is not going to put people in danger,” Julien says. 

In other words, while the AI is a capable assistant, the engineer remains the responsible professional.

AI becomes the orchestrator

Bentley’s portfolio spans structural analysis, geotechnical modelling and evacuation simulations. An AI agent can draw on all of these simultaneously, running checks across multiple engineering domains in parallel. 

Julian expands: “For an engineer, it’s very useful to be able to use a team of AI agents that can access the context about the project and turn design intent into AI recommendations – but with all of this validated so that the engineer can feel confident they were in control.”

One of the most time-consuming tasks in engineering is the production of technical drawings – detailed documents that communicate design decisions to contractors. Engineers often spend 30-50% of their time on a project annotating and detailing them, according to Julien.

AI is able to play a major role in automating much of this process. Bentley has trained models on engineering drawings to enable AI to generate annotations and measurements automatically. 

“The engineers just need to validate the drawing that has been generated by AI and make a few adjustments,” adds Julien. “They don’t need to do the measurements one by one.”

Bentley also allows firms to retrain these models on their own past drawings, so the AI learns a specific company’s house style while keeping the data private. 

“When you use your data to train AI models, those AI models remain only for you," says Julien. “They are not shared with anybody else and you’re in control of where they are being used.”

Building lasting infrastructure

A longer-term challenge looms in that infrastructure built today must remain accessible for decades – long after the software used to design it may have changed. 

Julien is a vocal advocate for open standards – agreed, publicly-available formats for representing data – as the solution.

He explains: “When you are creating a 3D model of a bridge, the format used to describe that model should be something that my kids and grandkids will be able to understand because it is an open standard.”

However, standards alone are not enough. Engineers also need open APIs that allow them to extract and move their own information freely.

Julien goes further, arguing that AI agents must be able to connect to infrastructure data through open protocols: “Open source, open APIs and open standards are the three legs of a truly open approach. For infrastructure, where a lot of it is commissioned by public money, this should be a strong mandate from governments and policy makers.”

The engineer of the future

Asked what becomes of the civil engineer as AI takes on more routine work, Julien states: “I really see the role of civil engineers becoming an orchestrator, where you create a team of AI agents doing specific parts of the workflow and you remain the one taking the important decisions.”

The pipeline of engineers coming out of universities is already insufficient to meet demand. AI, in this context, is a means of bridging that gap. 

“Engineers will be able to do much more meaningful and rewarding work, applying their human intelligence to infrastructure decisions for more impactful outcomes,” Julien concludes. “I don’t see AI replacing engineers. I see it helping them do a more rewarding job.”

To read the latest issue of AI Magazine, click here.