If the specialist keeps a parallel spreadsheet, you have this problem

The problem

If the rotating equipment specialist keeps a parallel spreadsheet, the problem exists.

When an overall level rises, the system paints it on a screen a specialist looks at when they can. For that to become work, somebody has to key a notification in by hand, copying across the diagnosis. Often it does not happen. And often it happens late. There is a symptom that gives this problem away in any plant: the rotating equipment specialist keeps their own parallel spreadsheet so as not to lose the thread between what the vibration system says and what is open in maintenance management. If that spreadsheet exists, the problem exists. The cost is double keying, latency between detection and action, and broken traceability precisely in the chain that matters most: symptom → diagnosis → intervention → outcome.

• The plant has already invested: sensors on critical rotating equipment, vibration routes, oil analysis, thermography. The data exists and it is good.
• It lives in the vibration system software, on a workshop PC, while maintenance management lives in the ERP. When an overall level rises, the system paints it on a screen a specialist looks at when they can.
• For that to become work, somebody has to key a notification in by hand, copying across the diagnosis. Often it does not happen. And often it happens late.
• The cost is double keying, latency between detection and action, and broken traceability in the chain that matters most: symptom → diagnosis → intervention → outcome.

How it fits the IRIS system

Connect in integration mode — two modern systems stitched through their APIs, neither replaced.

Connect in integration mode between two modern systems.

The loop closes in both directions: the threshold creates the notification with the tag, the criticality, the trend attached and the recommendation; and when the work order closes, the outcome goes back to the condition history. That closed loop is the raw material for a predictive plan that is actually tuned on this plant's equipment — on the main air compression train and the non-redundant pumps, where a missed week is the difference between a planned intervention and a trip.

• It reads events and trends from the condition system through its API.
• When a threshold is crossed, it creates the CMMS notification with tag, criticality, attached trend and associated recommendation.
• In the opposite direction, when the work order closes, it returns the outcome to the condition history.
• That closed symptom-diagnosis-intervention-outcome loop is the raw material for tuning a real predictive plan. Without replacing either system.

See the full IRIS architecture →

Before and after

The vibration alarm today versus the alarm wired into the CMMS

Aspect	Today	With iLEAN Connect
Threshold crossed	Painted on a workshop screen	A notification created, with context
Diagnosis	Copied by hand, if at all	Attached to the notification
Detection to action	Days	Minutes
Outcome of the intervention	Lost to the condition history	Returned when the order closes
Specialist's parallel spreadsheet	Necessary	Unnecessary
Either system	—	Neither is replaced

from on-screen detection plus manual keying — or none — to a notification created with context the moment the threshold is crossed. From the specialist's parallel spreadsheet to a single traceable history.

Impact estimate

Estimated impact — to validate with your own numbers.

The block below is an estimate to be validated against your plant's actual data. We put it forward so the committee has an order of magnitude; we refine it during the assessment.

• Estimated payback 5-10 months, with a disproportionate impact on the equipment availability indicator.
• What shrinks is the window between detecting and acting on non-redundant critical rotating equipment — in a single train, precisely the interval where unplanned downtime is decided.
• From on-screen detection plus manual keying — or none — to a notification created the moment the threshold is crossed.
• From the specialist's parallel spreadsheet to a single traceable history.

estimated payback 5-10 months, with a disproportionate impact on the equipment availability indicator. What shrinks is the window between detecting and acting on non-redundant critical rotating equipment, which in a single train is precisely the interval where unplanned downtime is decided. *Estimate to validate*.

And the fair question from the production manager

«Isn't a classic integration between the two vendors cleaner?» — it is, the day the project reaches the top of the IT list, and that is the problem: two modern systems still unstitched because there was never a window. This integration leans on the interfaces both already have, replaces neither, and asks for no shutdown. A notification is created only on a threshold the specialist defined, never on model judgment; where the model does intervene — turning the diagnosis text and the trend into the CMMS's notification fields — it is an anchored task, where the best models drop below 1.5% error [1], and the specialist sees the notification before it is released to planning.

^[1] OpenAI paper "Why Language Models Hallucinate", 2025 — on the reliability of AI in anchored tasks.