The aggregate quality Excel flies to ERP

The problem

The real batch data and the official batch data live in different places.

Two parallel worlds: the real spreadsheet where every batch's live data lives, and the official ERP, with data transcribed weeks late by someone who spends hours a month copy-pasting and making mistakes.

• The quality manager logs every recycled aggregate batch in a spreadsheet: sieve-by-sieve gradation, contaminant percentage, moisture. It is the document the whole plant trusts.
• That spreadsheet is the truth. The ERP, the official system, receives the same data weeks later, typed by someone who spends hours a month copy-pasting from one window to another.
• Nobody checks the copy against the original. Copying errors go in silently, and the management certificate is issued against quality data nobody has cross-checked batch to batch.
• Meanwhile several people hold ERP licenses only to enter quality data — a cost the CFO pays every month for a typing job. Nobody budgets it as such, but it is there in the license renewal.

How it fits the IRIS system

Connect in listen-inward mode — it watches the shared folder, not the person.

Connect's 'listen inward' mode: it watches the shared folder, not the person, reads the delta when the file is saved, parses it with an anchored LLM, validates it on screen, and uploads it to the ERP via API.

The quality manager keeps the spreadsheet that works for them. Connect reads only what changed when the file is saved, shows it on screen for sign-off and pushes it to the ERP by API, tagged with the batch and the person who signed. The spreadsheet stops being a parallel world and becomes the front door of the ERP for quality data.

See the full IRIS architecture →

Before and after

The gradation sheet off to one side versus wired to the ERP

Aspect	Today	With iLEAN Connect
Lag to the ERP	Weeks	Zero after sign-off
Monthly hours copy-pasting	Several	None
A batch out of gradation limits	Seen when someone looks	Flagged on save
Certificate vs quality data	Not cross-checked	Batch-to-batch cross-check
Who needs a lab-module license	Everyone who types	Only the validating supervisor
The spreadsheet itself at a C&D recycling plant	—	Same file, same columns

Weeks of lag → zero latency. Hours a month transcribing → zero. A management certificate with uncross-checked quality data → automatic batch-to-batch cross-check.

Impact estimate

Impact estimate — to be validated against your license count.

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 batches per week and on how many people hold a quality-module seat today.
• CFO angle: typically a 50-80% cut in quality and lab-module ERP licenses, because only the validating supervisor still needs one.
• Weeks of lag and hours a month of transcription go to zero, and so do the copying errors that came with them.
• And the management certificate is issued against quality data automatically cross-checked batch to batch, so the client receives a document backed by the lab and not by a retyped number.

CFO angle — direct ERP license savings: if several people log in just to enter quality data, with Connect only the validating supervisor needs a license. Typically a 50-80% cut in quality/lab-module licenses. Estimated payback 4-9 months. *Estimate to validate*.

And the fair question from the production manager

“What if it reads the wrong sieve column?” — parsing a spreadsheet whose structure you defined is an anchored task, where the best models drop below 1.5% error [1]. Each value is also checked against the physical logic of a grading curve — the percentage passing cannot grow as the sieve gets smaller — and nothing is pushed to the ERP until the supervisor signs it off on screen. A batch that breaks the curve is shown in red, not posted.

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