The lab test spreadsheet that no longer lives on its own

The problem

The test lab and the ERP live weeks apart.

The lab's real spreadsheet and the official ERP are two parallel worlds, with weeks of lag from manual transcription and copy-paste errors.

• Every shift, the final-test lab fills in a spreadsheet: refrigerant circuit leak-tightness, washer vibration and spin noise, dielectric strength and ground continuity.
• That spreadsheet is the real record. The ERP gets a transcribed copy weeks later, with the copy-and-paste errors that come with it and the gaps nobody notices.
• When a distributor reports a noisy washer batch, the quality engineer opens the spreadsheet, not the ERP — because the ERP is not up to date.
• And several people keep an ERP quality-module license only to type in those results, month after month. Their time goes into typing, not into analyzing why a batch of washers is louder than the last one.

How it fits the IRIS system

Connect listening inward — it reads the file, not the person.

Connect "listens inward", reads the file's delta when it's saved, parses it with an anchored LLM, and after the quality supervisor validates it, pushes it into the ERP via API.

The lab keeps its spreadsheet exactly as it is. What changes is that the official system stops lagging behind it, and the licenses paid only to copy results into the ERP stop being necessary. The test result is in the ERP the same shift it was measured. A distributor complaint can then be answered with data from the ERP, by batch and by test.

See the full IRIS architecture →

Before and after

The test spreadsheet today versus the connected spreadsheet

Aspect	Today	With iLEAN Connect
Lag between lab and ERP	Weeks	Second zero after validation
Hours per month transcribing	Several, every month	Zero
Noisy washer batch complaint	Answered from the spreadsheet	Answered from the ERP
Spreadsheet format	Must be respected	Read as it is, even as it changes
Who validates before the push	Nobody, it is just copied	The quality supervisor
Quality-module licenses	One per person typing	Only the one who validates

Weeks of lag → zero seconds. Hours per month transcribing → zero.

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, to be checked against your current ERP license invoice.
• CFO angle: a typical 50-80% reduction in quality and lab-module licenses, because whoever logs in only to type results stops needing to.
• From weeks of lag to zero seconds between the test bench and the ERP, once the supervisor validates.
• Hours of transcription per month brought to zero, along with the copy errors that came with them. The test technician spends that time on the bench, not at the keyboard.

Typical 50-80% reduction in quality/lab-module licenses, estimated payback 4-9 months. *Estimate to validate*.

And the fair question from the production manager

"That spreadsheet has been edited by five people over the years — does it hold up?" — that is the expected case. Parsing does not rely on fixed cells: it interprets columns, units, merged cells and notes as they are, which is an anchored task where the best models drop below 1.5% error [1]. And nothing reaches the ERP until the quality supervisor validates it. A value outside its physical range, such as a negative leak rate, is flagged instead of pushed.

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