Agentic diagnostics
Diag
Diagnose and troubleshoot with confidence. Investigate, analyze and act on your assets with AI-powered insights.
Investigating MNT-4412
Read the local gauge at the 1CC-TE-0341 test connection.
The gauge agrees, about 104 °F
The gauge reads normal
01The diagnostic
From alarm to answer, with a human in the loop.
Press Investigate on any ticket. Millwright opens a workspace, pulls the evidence, and works the fault out loud, on a board your whole team can see.
01
Builds the case
The agent reads the ticket, finds the alarm behind it, and lays the trend, the working theory and what it has ruled out onto a board. The reasoning is something you can point at.
CCW supply header is warmer than its limit
1CC-TE-0341 · 104.6 °F, limit 104 °F · 7h 40m
Nobody has picked this up yet
1CC-TE-0341 · CCW supply header
Working theory: the reading has been over the limit since the alarm came in. Either the header really is warmer, or the element measuring it has drifted.
02
Knows when to ask
Telemetry cannot tell a drifting instrument from a drifting process. When the data runs out, the run stops and asks the person standing at the machine, with the right procedure already open.
Please read the local gauge at the 1CC-TE-0341 test connection. If it also says about 104 °F, the header really is warm. If it reads normal, the element has drifted.
03
Closes the loop
The new work order keeps its link to the alarm it resolves, the discounted theories stay on the record, and the investigation is one press from a drafted article for the next crew.
Recalibrate 1CC-TE-0341, CCW supply header element
Repair · raised by this investigation
CCW supply header is warmer than its limit
1CC-TE-0341 · 104.6 °F, limit 104 °F · 7h 40m
Nobody has picked this up yet
02Why now
Detection was never the bottleneck. Diagnosis is.
A decade of money went into anomaly detection, and it worked — most plants now flag more anomalies than they have engineers to read them. Every one still lands in a queue for somebody to work out by hand. That is the part this closes, and none of the four things that make it possible is a better model.
Every conclusion shows its evidence
The points it read, the document it cited, and the theory it ruled out — all on the record, all auditable.
Not a score, with no reasoning attached
The investigation is the work order
Raised from the board and linked to the alarm it answers, so a closed ticket can say what it closed.
Not a dashboard, beside a CMMS somebody retypes it into
Day one is read-only, on one system
That is the whole first step. It fits inside one engineer's authority, and it does not wait on a data programme.
Not eighteen months of ingestion before anybody sees a fault
It stops and asks when the data runs out
Telemetry cannot tell a drifting instrument from a drifting process. On that fault, asking is the only safe answer.
Not a model with no way to say it is out of its depth
See it running on your own plant.
We are onboarding a small number of operations teams. Three questions, and we will be in touch.