Engineering Intelligence: Building Trust in AI for Rail Infrastructure

The UK rail network is in a state of paradox. On one hand, it is undergoing a period of immense investment and transformation, with mammoth projects like HS2 and East West Rail promising to reshape the nation’s connectivity. On the other, the existing network is under unprecedented strain, battling aging assets, funding gaps, and record-high cancellations.

This creates a formidable challenge: how can the industry deliver ambitious new infrastructure while simultaneously enhancing the resilience and efficiency of the old, especially when facing a recognised shortage of skilled engineers?

The answer lies in a technological leap forward. AI represents a step change for rail infrastructure, offering a way to bridge the gap between demand and capacity. But this isn’t about generic AI. For an industry where precision is non-negotiable and the consequences of error are real, we need AI with a Ph.D. in engineering.

Beyond automation: AI for optimised outcomes

For years, the promise of AI has been focused on automation. In rail, however, its true potential extends far beyond simply automating workflows. It’s about fundamentally optimising decisions and enabling outcomes that were not previously possible.

Imagine shifting from a reactive maintenance cycle, which often leads to costly downtime, to a predictive one. AI-powered solutions can analyse data from sensors on tracks and rolling stock, identifying signs of wear and predicting potential failures weeks in advance. This isn’t a theoretical ambition; it’s delivering tangible results today, allowing maintenance to be scheduled proactively to improve safety, boost reliability, and reduce lifecycle costs.

This same approach can be applied to optimise complex traffic flows, simulate the impact of climate events to build more resilient networks, and run countless design iterations for new stations or lines in a fraction of the time it would take a human. This isn’t about replacing engineers; it’s about empowering them with smarter, proven tools to meet the monumental scale of the UK’s infrastructure goals.

The trust imperative: AI built on real-world context

As rail professionals increasingly rely on AI, those decisions must be trustworthy. Trustworthy AI is built on context, requiring four pillars of infrastructure intelligence: project context (intent and constraints), environmental context (above and below ground), institutional context (past project data), and engineering context (physical principles). Bentley software helps ensure AI models are informed by the right context, drawing on data from Bentley Infrastructure Cloud, insights from Seequent, and the logic embedded in Bentley Open applications.

A partnership for the AI era

For more than 40 years, Bentley has partnered with infrastructure professionals to enhance productivity through software. That partnership continues and deepens in the AI era. We see our users already finding novel and sometimes unexpected ways to use the APIs in our applications, integrating our technology with their own AI agents. We welcome this creativity and believe it is just the beginning.

To accelerate this momentum, we’re launching the Infrastructure AI Co-innovation Initiative. This is a commitment to partnering more closely with our users to shape the next generation of AI-enhanced workflows. Through this initiative, we will prioritise which APIs to evolve to better support AI use cases. We’ll also explore new commercial models that reflect the evolving balance between AI-driven and human-driven work.

A foundation of trust: Our commitment to data stewardship

Bentley recognises that innovation must be grounded on a foundation of trust. As users rely on Bentley Infrastructure Cloud to manage their engineering files and data, they must be confident that their information is protected.

We remain firmly committed to the principles of data stewardship, as first announced at the Year in Infrastructure 2023 conference. Our users are in control of their data and decide if and how it is used for AI training.

To uphold this, we have implemented strict governance around AI model training. Only data that has been explicitly licensed or purchased for this purpose is used for our general models. This includes contributions from accounts that have agreed to supply data for the benefit of the broader user community. Crucially, users also have the option to fine-tune Bentley AI models with their own data for exclusive use within their own organisations.

To further support transparency, Bentley has introduced the Data Agreement Registry, an auditing system that provides clear visibility into how data was used to train specific Bentley AI models.

The challenges facing the UK rail sectors are significant, but so are the opportunities. By combining the irreplaceable experience of our engineers with the power of context-aware, trustworthy AI, we can deliver a safer, more resilient, and more efficient railway for generations to come. Learn more about Bentley’s rail and transit solutions here.