The Paint Lab Spreadsheet Flows to ERP

The problem

The paint shop's real quality data lives in a spreadsheet the ERP never sees.

Two parallel worlds coexist: the real paint-lab spreadsheet (where the live data lives) and the official ERP (data transcribed with a days-long lag by someone spending hours/week copy-pasting and making mistakes). The IATF 16949 auditor finds quality data that was never cross-checked between the two worlds.

• Every shift, the paint lab fills a spreadsheet with the control parameters of each batch: e-coat bath solids and conductivity, topcoat viscosity, film build measured on the body, appearance readings per color.
• That spreadsheet is the live data. The ERP gets it days later, copied by someone who spends hours a week retyping and occasionally shifting a row.
• Between the two worlds, decisions on the paint line are made with numbers the official system does not hold yet.
• And on IATF 16949 audit day, the auditor asks for paint quality data for a period and finds two versions that nobody ever reconciled.

How it fits the IRIS system

Connect listening inward — the folder is watched, not the technician.

Connect 'listen inward': it watches the lab's shared folder (not the person), reads the delta when the file is saved, parses it with a grounded LLM, shows the validation on screen to the quality manager, and after confirmation pushes it to the ERP via API.

The lab keeps the spreadsheet it trusts. Connect reads only what changed when the file is saved, shows the quality manager what is about to enter the ERP, and pushes it by API after confirmation. Retyping disappears, and so do the licenses that existed only to retype. The paint engineer sees the e-coat bath trend the same morning, not the following week, while there is still time to correct it.

See the full IRIS architecture →

Before and after

Paint lab data retyped by hand versus pushed from the folder

Aspect	Today	With iLEAN Connect
Lag from lab bench to ERP	Days	Zero, on save
Hours spent retyping	Several per week	None
Shifted row or typo	Enters looking like valid data	Caught against plausible ranges
E-coat and topcoat readings by batch	Two versions to reconcile	One validated record
Quality-module licenses	Paid for to key in data	Typically 50-80% fewer
Paint data for the IATF auditor	Reconciled by hand before the visit	Already consistent

A days-long gap between real data and ERP data → zero second. Hours/week 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, before counting the license saving.
• A typical 50-80% reduction in quality and lab-module licenses, because the people who only opened the ERP to type no longer need access.
• From a days-long gap between lab and ERP to zero, and from hours of transcription a week to none.
• And the paint quality data the auditor sees is the data the line actually ran with, batch by batch and color by color.

Estimated payback 4-9 months plus a typical 50-80% reduction in quality/lab-module licenses. *Figure to be validated*. (Payback 4-9 months · -50/-80% quality-module licenses)

And the fair question from the production manager

“What if it parses the wrong column and puts a viscosity where a film build should be?” — parsing a spreadsheet against your lab's own template is an anchored task, where the best models drop below 1.5% error [1]. And each value is checked against the plausible range for that parameter before it is shown to the quality manager, who confirms before anything reaches the ERP.

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