The zinc thickness you actually measure and the one your ERP shows are not the same number

The problem

Two parallel worlds, and sales promises specifications by looking at the wrong one.

There are two parallel worlds. The real spreadsheet, in a shared folder, where the actually measured zinc thickness sits. And the official ERP, transcribed weeks later by somebody who spends several hours a month copying and pasting, and introducing errors. When those two worlds diverge, the divergence is not discovered in a meeting: it is discovered in the customer's press, in the form of a claim. And there is a second, quieter, recurring cost: the ERP is licensed per active user, so every person in the laboratory who logs in purely to enter test results is consuming a full annual license to act as a keyboard.

• The real spreadsheet, in a shared folder, holds the zinc coating thickness and hardness actually measured coil by coil. The official ERP holds a version transcribed weeks later by somebody copying and pasting.
• Sales promises tolerances by looking at the ERP. When the two worlds diverge, the divergence is discovered in the customer's press, as a claim.
• Several hours a month of a quality technician go into transcription, and every paste is a chance to attach a hardness value or a coating thickness to the wrong heat, which is the error that surfaces months later and cannot be traced back.
• And a quieter cost: the ERP is licensed per active user, so every laboratory person who logs in purely to enter results consumes a full annual license to act as a keyboard.

How it fits the IRIS system

Connect in inward-listening mode — the folder, not the person.

Connect in listening-inwards mode. The system watches the shared folder — the folder, not the person — reads the delta when the file is saved and parses it with a grounded LLM. It presents the summary on screen and, once the quality lead validates, writes it to the ERP through the API, anchored to the heat number and the coil number. The laboratory keeps working exactly as before.

The laboratory keeps working exactly as before, in the sheet it has today. What changes is that the moment the file is saved, the new results are read, shown to the quality lead and, once validated, written to the ERP against the heat and coil number. The gap between what was measured and what the ERP shows goes to zero, and sales starts promising the zinc thickness that was actually measured.

See the full IRIS architecture →

Before and after

Today's laboratory versus the connected laboratory

Aspect	Today	With iLEAN Connect
Gap between measurement and ERP	Weeks	Second zero
Manual transcription	Several hours a month	Zero
Zinc thickness sold versus measured	Two different numbers	One number, heat by heat
Result attached to the wrong heat	Found in the customer's press	Cross-checked before the push
Sheet format	Has to be respected	Tolerated as is, even as it changes
Quality module licenses	Typically four	One, the lead who validates

from weeks of lag to second zero; from several hours a month transcribing to none; from an auditor holding laboratory data that was never cross-checked, to automatic matching heat by heat.

Impact estimate

Impact estimate — to be validated against your annual ERP invoice.

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, including the operational saving.
• 💰 CFO angle: if four laboratory and quality people log into the ERP only to enter tests, that is four licenses; with Connect only the validating lead logs in. 50-80% reduction in quality/laboratory module licenses, recurring every year.
• Several hours a month of an expensive technician recovered.
• And an auditor who finds laboratory data matched automatically heat by heat against the mill certificate, instead of a spreadsheet nobody cross-checked and an ERP that disagrees with it.

CFO angle — if the laboratory and quality team logs into the ERP only to enter tests, each person consumes a license. With Connect only the validating lead needs to log in: typically from 4 licenses to 1, a 50-80% reduction in quality/laboratory module licenses, recurring every year. Estimated payback 4-9 months including the operational saving. *An estimate to be validated* against your annual ERP invoice.

And the fair question from the production manager

“That spreadsheet has been modified by three technicians over ten years — does it hold up?” — that is the anticipated case: parsing does not depend on fixed positions, it interprets columns, units, merged cells and margin notes without asking anyone to standardize anything. It is an anchored task, where the best models drop below 1.5% error [1], and the quality lead validates on screen before anything is pushed to the ERP.

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