A photo of the Work Order at line startup

The problem

The batch that went into each line is rebuilt at the end of the day.

The Work Order and startup checklist — model, compressor or motor batch, reference torque — are printed from the ERP, but any shift-change note is handwritten and can take hours to get digitized, if it ever does before the shift ends. By day's end, someone has to reconstruct which batch went into which line.

• The Work Order prints from the ERP with the model, the compressor or motor batch and the reference torque for the door hinges and the drum bearing. That part is clean, because the ERP generated it.
• The trouble starts at shift change: a compressor pallet swapped mid-run, a torque adjusted on a screwdriver, a door-gasket reel replaced, a control board batch that arrived late. All of it is written by hand on the startup checklist.
• Those notes take four to eight hours to be typed in, if they are typed before the shift ends at all. Over a weekend, the gap grows to two days.
• By the end of the day someone has to reconstruct which compressor batch went into which washer or refrigerator line — exactly the record a warranty return will ask for months later, when nobody remembers the shift.

How it fits the IRIS system

Connect in photo mode — the line lead keeps the printed Work Order.

Connect photo mode recognizes the document with an anchored LLM, extracts model, critical-component batch and torque, and stores it in central memory. Before it crosses into the system, the supervisor confirms it with two taps on the line tablet (early human verification).

Nothing changes in the startup routine except a phone photo of two papers that already exist. That is why it is usually the first case a white-goods plant approves: it asks the line for nothing, it needs no integration on day one, and it gives the quality team the batch per line from minute one instead of at the end of the shift.

See the full IRIS architecture →

Before and after

Line startup on paper versus line startup on record

Aspect	Today, on paper	With iLEAN Connect
Digitization latency	4-8 hours, or never	Second zero after the two taps
Compressor or motor batch per line	Rebuilt at end of day	Queryable by batch on the web
Shift-change note	Handwritten in the margin	Extracted and tied to the Work Order
Reference torque used	On the printout only	Stored with the batch
Warranty return months later	Search through folders	One lookup by serial range
Line lead's routine	—	Same papers, one photo more

4-8 hours of digitization latency → zero seconds. Manual end-of-shift reconstruction → instant web lookup by batch.

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, depending on the number of lines and shifts you run.
• From 4-8 hours of digitization latency down to zero seconds once the supervisor confirms the summary on the line tablet.
• The manual end-of-shift reconstruction of which batch went where becomes an instant web lookup by batch, available to quality, planning and after-sales.
• Hours of supervisor and quality time recovered every week, line by line — time that today goes into copying papers instead of solving the line's problems.

Estimated payback 4-9 months depending on line/shift volume. *Estimate to validate*.

And the fair question from the production manager

"What if it reads the compressor batch wrong?" — pulling model, batch and torque from a Work Order with a known layout is an anchored task, where the best models drop below 1.5% error [1]. And there is a second net: before anything crosses into the system, the supervisor sees the summary on the line tablet and confirms or corrects it with two taps. A wrong batch never reaches the record unseen.

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