Delivery note, no typing

The problem

The truck waits, the clock runs and the batch numbers get rushed.

the truck waits racking up demurrage while the receiving supervisor types into the ERP. If they rush to avoid delay, some batches get logged wrong and surface weeks later in a claim that's nearly impossible to trace precisely.

• The truck arrives with vegetable oil, vinegar, egg yolk, glass jars, caps and cartons. Each delivery note carries dozens of lines and each pallet its own supplier batch.
• The receiving supervisor types it all into the ERP while the truck waits and demurrage starts to run.
• Rushing to free the dock means some batches get logged wrong, and nothing flags it at the time.
• The error surfaces weeks later in a claim, when you need to know which oil lot went into which mayonnaise batch and the receiving record says something else.

How it fits the IRIS system

Connect in photo mode for analog documents — the note as it comes, the labels as they are.

Connect, photo mode for analog documents. Vision AI reads the supplier's delivery note format as-is, and decodes GS1-128, QR, barcode or plain-text OCR on pallet labels. One verification tap on an industrial tablet and everything crosses to the ERP.

The supplier changes nothing. The supervisor photographs the note and the pallet labels, vision reads each format as it is and decodes GS1-128, QR, barcode or plain text, and one tap on the tablet sends it all to the ERP. The truck leaves and the batches are right. And demurrage stops being a line item nobody can explain. Each ingredient lot enters the ERP with its supplier batch, quantity and best-before date, ready for the first mayonnaise or ketchup batch that consumes it.

See the full IRIS architecture →

Before and after

Receiving at the dock today versus receiving by photo

Aspect	Today	With iLEAN Connect
Time per delivery	About 45 minutes	About 5 minutes
Supplier batch on each pallet	Typed under pressure	Read from the label
Delivery note format	Each supplier its own	Read as it comes
A batch logged wrong	Hidden until a claim	Zero: note and label cross-checked
Tracing an oil lot to a mayonnaise batch	Doubtful	Exact
Seats in the ERP receiving module	Several	Only the supervisor who verifies

45-minute receiving process → 5 minutes. Hidden batch errors → zero.

Impact estimate

Impact estimate — to be validated with your receiving volume.

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 3-8 months.
• CFO angle: the ERP receiving module typically needs several active seats; with Connect only the supervisor verifies with one tap, a typical 50-70% cut in receiving and warehouse seats.
• Receiving drops from 45 minutes to 5, and the truck stops racking up demurrage at your dock.
• And hidden batch errors go to zero, which is what keeps any later claim traceable down to the ingredient lot.

estimated payback 3-8 months. CFO angle: the ERP's receiving module typically requires several active seats; with Connect, only the supervisor verifies with one tap — a typical 50-70% cut in receiving/warehouse seats. *Estimate to validate.*

And the fair question from the production manager

"What if the label is creased or the note is a bad photocopy?" — reading a document and a label whose structure is known is an anchored task, and the best models stay under 1.5% error [1]. Codes are decoded, not guessed, and when the note says one batch and the pallet label another, the discrepancy is shown to the supervisor before the tap instead of being discovered later. The supervisor keeps the last word on every delivery. A blurred photo is simply asked for again, at the dock, while the truck is still there.

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