The quality spreadsheet that flows to the ERP on its own

The problem

Two parallel worlds and a person copying cells between them.

The real spreadsheet has live data; the official ERP has transcribed data days or weeks behind, entered by someone who spends hours a month copy-pasting and sometimes makes mistakes.

• The spreadsheet holds the live data: AOI and ICT yields per program, defect Pareto by part number, solder joint cross-sections, the first article inspection each customer asks for.
• The official ERP holds a transcribed version, days or weeks behind, entered by somebody who spends hours a month on copy and paste.
• Transcription errors surface at the worst possible moment: preparing a first article submission or answering an OEM customer's corrective action request, when the number in the ERP and the number in the sheet disagree in front of the customer.
• And because the sheet has been evolving for years across several hands, nobody will accept the standard answer, which is to ask quality to change how they work. The sheet wins that argument every time, because it is the only place where the numbers are actually right.

How it fits the IRIS system

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

Connect "inward listening": watches the lab's shared folder (not the person), reads the delta when the spreadsheet is saved, parses it with a grounded LLM, shows it for validation, and pushes it to the ERP via API.

The yield history a customer asks for at every quarterly review stops being rebuilt the night before. It is in the ERP, current, on the day they ask — which turns an afternoon of work into a query. And the lab keeps the sheet it knows, which is the only condition under which quality teams actually adopt this.

See the full IRIS architecture →

Before and after

Today's quality lab versus the connected lab

Aspect	Today	With iLEAN Connect
Gap between spreadsheet and ERP	Days or weeks	Zero-second
Manual transcription	Hours every month	Zero
Defect Pareto per customer program	Rebuilt on request	Already in the ERP
Copy-paste errors	Found during a customer review	Eliminated
Spreadsheet format	Has to be respected	Read as it is, even as it changes
Quality-module licenses	Several named users	One, the supervisor who validates

Days/weeks of lag between the real spreadsheet and the official ERP → zero-second. Hours per month transcribing → zero.

Impact estimate

Impact estimate — to be validated against your actual 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 several quality people log in purely to type results, those licenses go: −50 % to −80 %, recurring every year.
• Estimated payback 4-9 months counting only the license line and the transcription hours.
• From days or weeks of lag between the real spreadsheet and the official ERP down to zero-second, save after save.
• And yield history per customer program stops being rebuilt by hand ahead of every review, which is an afternoon recovered each time a customer asks and a number that no longer has to be defended twice.

−50 to −80% quality-module licenses · payback 4-9 months. Estimate to be validated.

And the fair question from the production manager

“Several people have been editing that sheet for years — does it really hold up?” — that is the anticipated case. Parsing does not depend on fixed cell positions: columns, units, merged cells and margin notes are interpreted without anyone standardizing anything. It is an anchored task where the best models drop below 1.5% error [1], and the quality supervisor validates before the push.

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