A single cause gets buried as new data streams in.
Your systems report accurate data, but the small signals are the first to get lost as more events arrive. An electrical event can affect downstream cooling equipment, and that equipment can trigger a mechanical alarm hours later. The BMS reports the mechanical fault, while the electrical event behind it sits in another system with nothing to link the two.
The operator ends up investigating a symptom. Root-cause identification is delayed, related issues keep occurring, and SLA risk grows. The cost is downtime, cascading failures, and delayed incident response. That costs the company money.
That is visibility without context, and it calls for operational triage.
This article separates visibility (seeing the data) from operational context (understanding cause, consequence, and priority). It starts with what happens when alarms fire.
Each system generates its own information, often on its own screens or separate machines.
Even when everything opens in a browser, the operator jumps between systems to connect the readings. The systems themselves are connected, and an electrical event can ripple into the mechanical system. But the interfaces don’t show it.
Consider a simple case. An alarm fires in the mechanical systems. It monitors a single data point and says that a threshold was crossed. It doesn’t say why. The mechanical screen shows the symptoms, and the cause sits in the electrical system.
To investigate, the operator works backwards from the latest event. Data arrives with the most recent events first, so newer events keep pushing the original electrical event down the screen. New events keep arriving that the operator can’t ignore, and each one delays understanding. The most likely root cause gets lost under new streaming data, cross-system ripple effects, and other issues that surface during the analysis. Until it’s found, related issues keep occurring.
We think an operator’s hours belong on the cause, not the reconstruction.
Visibility tells an operator that something changed. It doesn’t tell them why.
In the simple case, the mechanical alarm is visible, and so is the electrical event behind it on a different screen. Context is the connection between the two, and neither screen shows it.
Each level lets the operator infer a little more, then stops short of the next.
Not every facility needs to reach level 5. Value begins with better context.
Operational context is the foundation of operational intelligence. It connects equipment, systems, history, and documentation. Operational intelligence uses those relationships to support root-cause investigation, prioritization, and recommendations.
So how does a facility build that context?
Prism builds a living knowledge graph of your facility. It maps physical locations, equipment connections, and manufacturer documentation.
Each machine is a node, and each connection records what it physically means. A chiller plant that cools specific data halls is mapped as a connection between that plant and those halls. Location works as a hierarchy, like an address tree that drills down from the facility to the asset.
Any one of these feeds reports a value. The relationships between them show what that value affects.
A facility changes, so the model has to keep up. The model starts with as-built construction documents and stays current as operating conditions, equipment changes, and other facility updates occur.
With the relationships mapped, the mechanical alarm stops being a mystery.
Root cause becomes something the operator can investigate. Prism runs that analysis in the background on every alarm and fault. The graph records that the mechanical equipment depends on the electrical equipment feeding it. The LLM uses that structured context to correlate signals across systems and surface likely causes for the operator to investigate.
We don't promise a single root cause every time, and the operator still makes the call.
One event can fire alarms across several systems. Prism pieces those alarms together and helps identify a likely common cause.
In one utility event incident, hundreds of alarms came in. Prism traced them to a common cause. The team ran its own investigation in parallel and reached the same root cause in about 12 hours. Prism produced an impact statement in less than 15 minutes.
This isn’t a fault of the operators. It’s a result of separate systems, missing context, and clunky interfaces.
Alarms are classified by static priority, usually across four classifications. An alarm marked as critical will show up as critical every time it goes off, no matter what else is happening. A priority that never changes can’t reflect a facility that does.
When a utility event affects power and cooling while a chiller fault appears on another loop, both may carry the same static priority. The operator has to weigh current conditions, dependencies, and likely impact to decide which one threatens the SLA first. That is the same logic used in emergency-room triage: priority depends on the condition of the entire room, not arrival order.
Prism uses static priority as one input when triaging concurrent incidents. It weighs that priority against current facility conditions, equipment dependencies, and potential impact.
A likely cause still needs a check, and part of that check sits in the CMMS.
Prism reads CMMS records alongside live equipment telemetry. When an alarm or fault comes in, it checks recent work orders on the affected equipment. It flags recent maintenance as a likely cause, or as an explanation that lowers the priority of the finding.
The same connection answers a harder question: did the maintenance actually fix the problem? A completed work order records the work performed, but it may not show whether the repair resolved the issue. Confirming that can require comparing equipment performance before and after the work. Prism uses work order dates to set comparison windows, retrieves telemetry for both periods, and summarizes the changes. Trend charts mark when the maintenance took place.
During an investigation, Prism surfaces work orders, incidents, and change requests tied to the insight, and the operator can see them in the same view as the telemetry.
Prism only reads CMMS records. It doesn’t create, edit, or close work orders yet. Your CMMS keeps doing its job, and Prism adds maintenance context to investigations. Inconsistent failure codes or close-out notes stored as links outside the CMMS limit what Prism can see.
Prism detects faults and data drift continuously, so early signals reach the operator while a problem is still forming. A threshold alarm can't flag a failure building below the threshold.
Take a chiller plant. One chiller runs at lower efficiency than comparable equipment, and no alarm fires. Its performance has drifted from its historical baseline and from its peers, but no single reading crosses a threshold.
Prism catches that drift through continuous analytics across equipment data, operating history, and system dependencies. Here, it traced the pattern to a refrigerant circuit issue on one chiller, then to the fans behind it. The operator can start on the cause before related problems spread across systems.
When an event does run, Prism provides the analysis while it is ongoing, not days or weeks later. The operator fixes the underlying issue instead of the symptom, and leadership can get a root-cause analysis report in the same timeframe.
Prism is the connective tissue between your systems, and it doesn’t replace them.
It brings information into a single interface, backed by an intelligence layer that uses LLMs to synthesize data for operators. Prism turns that connected context into operational intelligence for data center operators.
Your BMS, EPMS, DCIM, CMMS, and equipment controllers keep doing their jobs. Your operators keep making the decisions. Prism gives them better information instantly.
Context comes first, and agents come second. Reliable automation starts with trusted data, accurate equipment relationships, and a clear understanding of operational constraints. AI agents take on defined tasks, operators keep oversight, and automation increases only within defined boundaries and approved policies.
Further reading: Is your data ready for AI? A guide to data readiness
When the mechanical alarm fires, the operator should see the electrical event behind it and the history of both systems in one place. That's the standard we hold Prism to.
To see how Prism connects systems, equipment, and history around what you already run, request a demo.
Subject matter expert: Christopher Vause | Chief Product Officer, Phaidra