Lab analysis reaches the ERP without anyone typing it

The problem

The freshest data in the plant, and the ERP finds out weeks later.

The lab of a plant with raw milk intake, a UHT line, a juice line and a cheese line generates a volume of analysis no shared spreadsheet survives with dignity. There are versions. There are local copies someone took away to work without a lock. There are cells formatted as text where a number should be. And the ERP, where batch release actually lives, requires all of it to be transcribed by hand. Every month several person-days go into typing what was already written. As a result, the ERP data runs systematically weeks behind reality, and out-of-specification results are caught in the monthly review instead of when they happen.

• The lab sheet holds the intake analysis, alkaline phosphatase, pH, degrees Brix, acidity and somatic cells, shift by shift. It is the freshest data in the plant.
• It is a shared spreadsheet with versions, local copies someone took away to work without a lock, and cells formatted as text where a number should be.
• And the ERP, where batch release actually lives, requires all of it to be transcribed by hand. Several person-days a month go into typing what was already written.
• As a result the ERP runs systematically weeks behind reality, and an out-of-specification result is caught in the monthly review instead of when it happens.

How it fits the IRIS system

Connect in inward listening mode — the file, not the person.

Connect in "listen inward" mode. The flow:

The lab does not change tools or learn a new one. They keep their sheet, with all the formatting quirks accumulated over years, and iLEAN reads it the moment it is saved. What disappears is the transcription — and the licenses that existed only to do it. An out-of-specification acidity or phosphatase result is flagged when the file is saved, before anyone has looked, and the quality manager validates on screen before the API posts anything to the ERP.

• iLEAN observes the shared folder and reads the file when it is saved. This deserves precision: it does not monitor any person, it reads a file.
• A model grounded in that specific sheet's structure extracts the values, even with the formatting quirks accumulated over years.
• Each value is linked to the corresponding batch and line in central memory.
• iLEAN immediately flags anything out of specification, before anyone looks.
• The quality manager validates on screen and, after that validation, the API posts the data into the ERP's quality module.

The lab does not change tools or learn a new one. They keep their sheet. What disappears is the transcription.

See the full IRIS architecture →

Before and after

Today's lab versus the connected lab

Aspect	Today	With iLEAN Connect
Gap between real result and ERP	Weeks	Minutes
Manual transcription	Several person-days a month	Zero
Out-of-specification result	Caught in the monthly review	Flagged when the file is saved
Sheet format	Would have to be standardized	Read as it is, quirks included
Who validates before posting	Nobody, it is typed	The quality manager, on screen
Licenses on the quality module	Six	One, the person who validates

from weeks of lag between real data and ERP data, to minutes. From catching out-of-specification results in the monthly review, to catching them when the file is saved. From several person-days a month of transcription, to zero.

Impact estimate

Estimated impact — to validate against your actual 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.

• 💰 CFO angle: the ERP bills per named user. If six lab people log in purely to enter results, that is six licenses on the quality module: from 6 to 1, a 50-80 % reduction, recurring every year.
• Estimated payback 3-6 months counting licenses and transcription alone.
• Several person-days a month of lab time recovered.
• And out-of-specification results caught when they happen instead of weeks later, which is what protects the batch release.

direct CFO angle. If six lab people log into the ERP purely to enter results, that is six licenses on the quality module. With Connect only the person who validates logs in: from 6 to 1. At the usual rates of a high-end corporate ERP, that is a recurring five-figure annual saving. Typical ratio for this archetype: −50 % to −80 % on lab/quality module licenses. Estimated payback 3 to 6 months counting licenses and transcription alone. *Estimate to validate against the actual ERP invoice.*

And the fair question from the production manager

«That sheet has been edited by six people for ten years — does it hold up?» — that is the anticipated case: parsing does not depend on fixed cell 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 manager validates on screen before anything is posted to the ERP.

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