The rounds stay on paper; the data does not

The problem

The only proof the equipment is fit to start lives in a binder.

Preservation and preventive rounds are long routes across a large site, with stops at dozens of pieces of equipment. At each stop, things get written down that are, literally, the only evidence that the equipment is fit for service: shaft rotation, nitrogen blanketing pressure, oil level and condition, motor space heater, coating condition. That sheet ends up in a tray, then in a binder, then in an archive nobody ever opens again. And here is the real cost, which is not the writing time: nobody is going to read three hundred sheets to discover that an oil level has been drifting down for six weeks. The trend is written down, in the plant, right now. Nobody can see it. The aggravating factor appears in any asset that spends a long period in preservation awaiting start-up: that paper history is the documentary proof that the equipment is fit to start. If it is not queryable, it is not proof.

• A preservation round stops at dozens of tags: shaft rotation, nitrogen blanketing pressure, oil level and condition, motor space heater, coating. Each reading is the evidence that the equipment is fit for service.
• The sheet goes to a tray, then a binder, then an archive. Nobody is going to read three hundred sheets to notice an oil level drifting down for six weeks.
• The trend is written down, in the plant, right now — and nobody can see it. That is the real cost, not the writing time.
• On an asset that spends months in preservation before start-up, that paper history is the documentary proof it is fit to start. If it cannot be queried, it is not proof.

How it fits the IRIS system

Connect in photo mode — the technician keeps closing the sheet on paper.

Connect photo mode.

It is the case with the least friction in the whole matrix, because it asks nothing of the field: the same sheet, the same pen, the same gloves, plus one photo. That is usually why it is the first one a maintenance manager approves — and the one that turns the paper archive into a condition baseline for the asset. Nobody in the field has to learn a new screen.

• The technician closes the sheet as always and photographs it with a phone or industrial tablet.
• A language model anchored to the known structure of that sheet extracts the equipment tag, every numeric reading with its measurement point, the ticked boxes, the handwritten remarks and the signature.
• The result lands as a visual summary for early human verification (see case 5).
• Once validated, the data enters central memory tied to the equipment and exists from then on as a time series queryable in plain language. Zero habit change and zero typing in the field.

See the full IRIS architecture →

Before and after

The round on paper versus the round captured

Aspect	Today, on paper	With iLEAN Connect
Data latency	Days or weeks — or never	Second zero
Oil level drifting for six weeks	Invisible in a binder	A curve per measurement point
Who took the reading	A signature on a sheet	Tied to the tag, with name and time
Proof the equipment is fit to start	Three hundred sheets in an archive	One query by tag
Historical archive	Dead weight	Condition baseline, digitized retroactively
Technician's habit	—	Unchanged: one photo

from days or weeks of document lag — or never being digitized at all — to second zero. From "we don't know" to a curve per measurement point, with who took each reading. From a paper archive as dead weight to a condition baseline for the asset.

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 3-8 months, depending on rounds per month and points per route.
• Transcription and document-hunting time recovered by the planning team.
• From days or weeks of document lag — or never being digitized — down to second zero.
• A frequent bonus: digitizing the historical archive retroactively turns dead binders into the equipment condition baseline.

estimated payback 3-8 months depending on rounds per month and points per route; recovers the planning team's transcription and document-hunting time. *Estimate to validate*. A frequent bonus: retroactively digitizing the historical archive turns dead binders into the equipment condition baseline.

And the fair question from the production manager

«What if it misreads a pressure or an oil level on a handwritten sheet?» — reading fields from a document whose structure is known is an anchored task, where the best models drop below 1.5% error [1]. And there is a second net, built in: the summary lands on the technician's tablet at the equipment and they confirm or correct before anything enters central memory. A value outside the physical range of that tag — a blanketing pressure the regulator cannot deliver — is flagged, not accepted, and every correction teaches the system how this plant writes its sheets.

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