A grid operator rarely lacks data. Measurements sit in operational systems, asset locations in GIS, maintenance history in enterprise systems, and defects in inspection records. The difficulty is bringing those records together when someone needs to make a decision.
Consider a transformer showing an unusual condition trend. The measurement alone cannot tell an engineer how urgent the issue is. They may also need to know which line it feeds, what an inspection found, whether related assets show similar symptoms, what work is already planned, and which operating procedure applies. Today, a skilled person often assembles that picture across several systems.
This is a challenge for AI in grid operations. A language model can retrieve documents and summarise records, but it cannot safely infer that two identifiers refer to the same physical asset. Nor should it decide who is authorised to act. Those relationships and rules need to be made explicit.
A useful starting point is one operational question: which assets should an engineer review this week, and why?Answering it well requires four things:
Asset identity. Connect records from different systems to the correct asset, while flagging uncertain matches for review.
Operational relationships. Show how assets connect to the network, inspections, defects and work orders.
Traceable evidence. Let an engineer inspect the source and timing of each observation behind a recommendation.
Clear authority. Keep recommendations, approvals and operational actions within the organisation’s established procedures.
This approach can begin with a focused use case. An operator can define the decision, connect the necessary sources and test the resulting view with the people who make that decision. The asset context developed for that first use case can then support others, such as inspection follow-up or maintenance prioritisation.
At NeuroCluster, we are developing GridSight around this principle: connecting operational, spatial, asset and inspection information so grid teams can examine a decision with its evidence and controls in view. Our published product examples illustrate the intended workflow; results would need to be measured in an operator’s own environment.
The question for grid leaders is therefore practical: which recurring decision still requires an experienced colleague to reconcile several systems by hand? That is a strong place to start making AI useful in operations.
Christiaan van Steenbergen is Founder & CTO of NeuroCluster, based in Brainport Eindhoven. Learn more at https://www.neurocluster.com/gridsight