The serial marker that no longer gets hand-typed

The problem

The last physical step before the customer still depends on hand typing.

Serial marking is the last physical step before the customer — a typing error on a lot change breaks full warranty traceability, the worst case when the customer claims warranty by manufacturing date.

• The marker that prints serial number, date code and plant on every battery is fairly modern and already has an API, but it sits isolated from the ERP. Nobody connected the two because each was bought for a different purpose at a different time.
• At every lot or model change, an operator types the new values by hand at the marker's console. Lot changes happen several times a shift on mixed lines, and each one is a chance to mistype.
• One digit wrong in the date code breaks warranty traceability for every battery in that run. The mistake is invisible on the line: the code looks correct, it just belongs to another lot. On mixed EFB and AGM lines, where model changes are frequent, that is where most marking mistakes start.
• That is the worst case when the customer or the OEM claims warranty by manufacturing date: the batteries in the field point to the wrong lot. Containment then has to be drawn far wider than the actual problem, because the date code cannot be trusted.

How it fits the IRIS system

Connect stitching two modern systems via API — the ERP talks to the marker.

Connect stitches two modern systems together via API. When the ERP flags a program start, it automatically publishes lot + date + model to the marker. The operator no longer types anything.

This is the quickest case to deploy because both ends already speak API. When the ERP flags a program start, lot, date and model reach the marker on their own, and the operator stops being the weakest link in warranty traceability. It is a small integration with a large effect on the one record the customer reads on every battery.

See the full IRIS architecture →

Before and after

Serial marking typed by hand versus published by the ERP

Aspect	Today	With iLEAN Connect
Date code at a lot change	Typed at the marker console	Published by the ERP via API
Human-origin marking errors	Occasional, with warranty impact	Zero
Warranty claim by manufacturing date	May point to the wrong lot	Points to the right lot
Operator task at a model change	Retype serial, date and plant	One tap to confirm the start
Marker and ERP	Two isolated systems	Stitched together
Deployment effort	—	Short: both ends already have an API

Occasional typing errors with warranty-traceability impact → zero human-origin error in marking.

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 3-6 months, a fast one.
• The return comes from a drastic reduction in the risk of a poorly traced warranty claim. One avoided claim with an untraceable date code usually covers the whole deployment.
• Zero human-origin errors in serial and date code marking. The operator's time at each lot change goes back to the line.
• And when a claim arrives by manufacturing date, the date code on the battery leads straight to the right lot. Which keeps a warranty containment limited to the lot that actually failed, not the whole month. That shortens every warranty investigation that starts from a battery returned by a dealer.

Fast payback, 3-6 months, from the drastic reduction in the risk of a poorly-traced warranty claim. *Figure to validate*.

And the fair question from the production manager

“If it is just an API link, where does the AI come in?” — in the check, not the typing. Before publishing, iLEAN confirms that the lot, date and model match the active program in the ERP; that comparison is an anchored task, where the best models drop below 1.5% error [1]. If anything disagrees, the marker is not updated and the operator is asked to confirm. Every publication is logged with the program and the time, so the trail is complete.

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