A paperless cutting order, at zero latency

The problem

The order that reaches the shear is not always the order the customer asked for.

In a steel service center, every cutting batch starts with a paper order listing customer, grade, thickness and dimensions. Any last-minute correction phoned in gets scribbled in the margin and rarely makes it back into any system. When a customer disputes a measurement, nobody can reconstruct which exact order reached the shear.

• Every cutting batch starts with a paper order: customer, grade, thickness, the format it comes out of and the sizes to cut. It is written at the counter and it walks to the shear in somebody's hand.
• A correction phoned in at the last minute — a width down from 2000 to 1990, one more piece, a grade swapped for whatever is actually on the rack — gets written in the margin and rarely makes it back into any system.
• Between the margin correction and the system, four to eight hours pass, and part of it never gets typed at all.
• So when a metal fabricator disputes a measurement three weeks later, nobody can reconstruct which exact order reached the shear, and the discussion is settled by whoever remembers harder. In a trade where the same customer orders three times a week, that conversation is expensive twice over.

How it fits the IRIS system

Connect in photo mode — one photo when the batch starts, and the order exists.

Connect photo mode: the operator photographs the order when starting the batch. An anchored LLM extracts customer, grade, thickness, dimensions and the handwritten note, cross-checking it against the ERP order. The data lands in central memory before cutting starts, with no change to the operator's routine.

The operator's routine does not change: they keep working from the same paper order and simply photograph it before the first cut. That is why this is usually the first case a steel warehouse approves, and why the order stops depending on a sheet of paper surviving the shift.

See the full IRIS architecture →

Before and after

Today's cutting order versus the captured order

Aspect	Today, on paper	With iLEAN Connect
When the order exists in a system	4-8 h later, or never	Zero latency
A correction phoned in	Written in the margin	Extracted and tied to the batch
A measurement dispute	Word against word	The exact order that reached the shear
Grade, thickness and format	Retyped from paper	Cross-checked against the ERP order
Sizes cut out of the format	Only on the paper	Recorded with the offcut left over
Operator's routine	—	Unchanged: one photo

4-8 hours of latency between the paper correction and its reflection in the system → zero latency.

Impact estimate

Impact estimate — to be validated with your 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 4-9 months, depending on how many cutting orders you put through per day.
• Counter and administrative transcription time recovered, order after order.
• From 4-8 hours of latency — and orders that never got typed at all — to zero latency on every one.
• And a dispute over a measurement stops being word against word: the exact order that reached the shear, corrections included, becomes a query.

Estimated payback of 4 to 9 months depending on daily order volume. Estimate to be validated.

And the fair question from the production manager

“What if it misreads a size in a handwritten correction?” — reading fields off a document with a known structure is an anchored task, where the best models drop below 1.5% error [1]. And there are two nets behind it: the summary goes to the tablet at the cutting station for the operator to confirm or correct, and grade, thickness and format are cross-checked against the ERP order and against what the warehouse actually holds, so a combination that does not exist is flagged instead of accepted.

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