One of the biggest challenges facing the rail industry today is not a lack of data. Quite the opposite.
Rail operators generate enormous volumes of operational information every day across signalling systems, infrastructure monitoring, maintenance records, controller logs, crew reports and operational databases. The challenge has never been collecting the data. The challenge has been bringing it together quickly enough to make better operational decisions.
That challenge is particularly acute in delay attribution, where investigators often need to piece together evidence from multiple systems, interpret free-text narratives, apply complex industry rules and reach consistent conclusions under significant time pressure.
Today we’re delighted to announce two important milestones.
Great Western Railway’s delay attribution solution has entered production.
And, following its successful deployment, we are launching the solution commercially as Geminos PathWay™ for Rail – Delay Attribution, the first member of our new family of purpose-built AI applications, at Rail Live 2026.
From Customer Solution to Commercial Product
The GWR project demonstrated something that many organisations have struggled to achieve with AI.
Not simply a proof of concept.
Not another chatbot.
A production-ready operational application that solves a real business problem.
The success of the deployment has enabled us to package the solution so that other railway operators can rapidly deploy the same capabilities, adapted to their own operating environment, data sources and business rules.
Rather than embarking on lengthy software projects, operators can now adopt a proven solution that has already demonstrated its value in a live railway environment.
Bringing the Evidence Together
Delay attribution is fundamentally a knowledge problem.
The information required to understand an incident rarely resides in a single application. Attribution teams routinely work across systems such as Sentinel, defect logs, CCIL and QUARTZ, manually assembling the evidence needed to determine what happened.
Much of the most valuable information isn’t stored in structured fields at all. It exists inside controller logs, maintenance reports, crew narratives and station observations.
Historically, extracting meaningful knowledge from this unstructured information has been difficult, expensive and inconsistent.
Geminos PathWay for Rail automatically consolidates all of this information into a single investigation workspace.
Instead of investigators spending their time searching for information, the platform assembles the relevant evidence automatically, allowing them to focus on understanding the incident itself.
Understanding What the Text Actually Means
Large language models have transformed our ability to understand free text.
But enterprise decision making requires much more than language models alone.
Railway operations involve specialised terminology, operational concepts, complex business rules and causal relationships that generic AI systems simply don’t understand.
This is where the Geminos Enterprise Knowledge Graph becomes critical.
Every PathWay solution is built on an Enterprise Knowledge Graph that captures the operational concepts, relationships, terminology and business rules specific to the customer’s industry.
Rather than simply identifying keywords, the system understands that a controller’s log, a defect report and a guard’s narrative may all describe different aspects of the same underlying event. It can connect those pieces of evidence, reconcile conflicting information and build a coherent understanding of what actually happened.
This ability to transform unstructured text into structured enterprise knowledge is becoming one of the most valuable applications of AI.
Every organisation possesses enormous amounts of knowledge locked inside reports, emails, maintenance records, inspection notes and operational logs. Until recently, that knowledge was largely inaccessible. Today, it can become the foundation for faster, better operational decisions.
From Knowledge to Better Decisions
Understanding the evidence is only the beginning.
The objective is improving decisions.
Once information has been extracted and connected through the Enterprise Knowledge Graph, it becomes possible to apply causal reasoning.
The platform evaluates evidence against operational context, timing, location and historical events before determining whether a proposed cause can be confirmed, ruled out or requires additional evidence.
Recommendations are generated within the framework of the Delay Attribution Principles and Rules (DAPR), which are represented directly within the knowledge graph. Rather than checking compliance after reaching a conclusion, the system reasons within the rules themselves.
The result is a transparent, explainable recommendation supported by a complete audit trail.
Looking Beyond Individual Incidents
Individual incidents rarely occur in isolation.
By connecting information across services, infrastructure, rolling stock, routes and historical events, Geminos PathWay can identify recurring operational patterns that would otherwise remain hidden.
An apparently isolated delay may actually be part of a recurring infrastructure issue, an emerging fleet reliability problem or an operational bottleneck affecting multiple services.
This broader perspective allows railway operators to move beyond reactive investigation towards continuous operational improvement.
Why We Can Deliver in Weeks, Not Years
Traditional enterprise AI projects often spend months integrating systems before delivering any business value.
Geminos takes a different approach.
Every PathWay solution is built using ArchWay™, our enterprise AI architecture, which combines Enterprise Knowledge Graphs, large language models, deterministic software, business rules and Causal AI into a single production-ready framework.
Rather than beginning with technology and searching for a use case, we begin with the operational decision that needs to be improved.
ArchWay allows us to rapidly assemble the knowledge graph, integrate operational data, interpret unstructured information and build causal models around the customer’s specific challenge.
The result is dramatically shorter time-to-value.
Customers see a working application built around their own use case within days. Production-ready systems follow within weeks.
A Blueprint for Operational AI
The initial GWR deployment focuses on TO and R8 attribution scenarios, representing approximately 40% of delay incidents.
But the significance extends far beyond delay attribution.
Geminos PathWay demonstrates a repeatable approach for solving operational problems that require organisations to combine fragmented information, understand large volumes of text and make complex, explainable decisions.
The same architecture can be applied across compliance, safety investigations, maintenance planning, asset management, operational performance, root cause analysis and risk management.
In fact, Geminos PathWay is being developed across multiple industries where organisations face exactly the same challenge: turning fragmented information into actionable knowledge.
The GWR deployment shows that when AI is grounded in enterprise knowledge, business rules and causal reasoning, organisations no longer need to choose between innovation and production.
They can have both.
And they can achieve it in weeks, not years.
See Geminos PathWay™ for Rail at Rail Live 2026
Geminos PathWay™ for Rail – Delay Attribution will be demonstrated publicly for the first time at Rail Live 2026.
If you’re attending the event, we’d love to show you how we’re helping railway operators move from fragmented operational data to faster, more consistent and more transparent delay attribution in weeks rather than years.
And if delay attribution isn’t your biggest challenge, let’s talk about what is.
The same PathWay approach can be applied across safety investigations, compliance, asset management, maintenance planning, root cause analysis and many other operational decision-making challenges.
Bring us your toughest operational problem. We’ll show you a working application built around your use case in days—and a production-ready solution in weeks.