A paper delivery note should not block the unloading of the truck.

The problem

The receiving bottleneck is not the truck: it is the keyboard.

Any given morning at a plant that makes food products:

• At 9:30 the truck comes in with a large order: high-gluten flour, seasoning, sesame, packaging film. Twelve pallets in total.
• The supplier hands over an A3 delivery note with twelve detail lines, each with its product, quantity, batch and expiry date. Each pallet also carries its own batch labels stuck on the side.
• The receiving lead has to sign the delivery note and log into the ERP to register line by line, batch by batch, before that flour can be booked as available for production.
• Do it slowly to avoid mistakes and the driver waits 45 minutes charging demurrage. The demurrage invoice is paid by the plant.
• Do it fast and there are batch errors that surface three weeks later in a health alert that can no longer be traced — because the batch the ERP says is not the batch that actually came in.
• Today's alternative: either an intern transcribing, or an ERP receiving module that costs a license per user plus implementation. And even then, someone keeps typing.

The bottleneck is not the truck, or the warehouse, or the ERP. It is the person who has to recontextualize by hand what is already written on a piece of paper.

How it fits the IRIS system

Connect in "photo of the analog" mode — digitization starts at the dock.

iLEAN does not add another screen for the receiving lead; it takes away the two they had. The piece is iLEAN Connect in its most characteristic mode: photo of the analog, the one that digitizes paper production sheets, delivery notes, labels and signatures without asking the supplier to change anything in their system.

The receiving lead pulls out the phone or the rugged tablet and takes two photos. Connect reads, extracts, proposes. The person validates with one tap. The ERP receives a clean entry. The truck leaves in 5 minutes, not 45.

• Photo of the whole delivery note. The camera recognizes the supplier's delivery note format without asking the supplier to change it. The inference understands the tabular structure and extracts what matters: delivery note number, lines (product, quantity, batch, expiry date), supplier, date. It works for the old-school flour mill's handwritten A3 and for the film supplier's printed A4 with a barcode.
• Photo of the pallet coming in. For the pallet's batch labels, the camera recognizes GS1-128, QR, barcode or, if the label only carries plain batch-and-date text, reads it via OCR. If three batches are mixed on one pallet, it separates them as three independent entries instead of inventing an averaged one.
• Everything goes to the AI central memory at second zero. It does not wait for the end of the shift, or for "someone to write it up properly". The raw information is stored with its original photo as evidence.
• One-tap validation. The receiving lead sees a summary on their tablet ("12 lines extracted, 3 unique batches, one doubtful field flagged in yellow — confirm?") and signs with one tap. If something does not match, they correct it before signing — never after.
• Automatic insertion into the ERP via API/RPC. Without anyone in receiving having opened the ERP's receiving module, the lines go in clean, with their batch, quantity and expiry date. The ERP's bulk-capture module is no longer needed.
• Result at the dock: the driver leaves in five minutes. Zero demurrage invoices. The flour is booked as available for production without waiting for the last keystroke.
• Direct savings on ERP licenses. The ERP's receiving module typically demands 2-3 active licenses (receiving lead + peak-hour support + supervisor). With Connect, only the receiving lead validates with one tap, without properly entering the ERP. Typical result: one single ERP receiving license instead of three, plus removal of the bulk-capture module cost. What used to be paid per user to "enter data" is now paid for by the phone's camera.

See the full IRIS architecture →

Before and after

Keyboard at the dock vs. photo to the central memory

Aspect	Manual receiving + ERP module	With iLEAN Connect (photo to the ERP)
Truck unloading time	45 min while someone types	5 min — the photo is processed afterwards
Receiving lead's work	Typing 12 lines into the ERP	One photo + one tap of validation
Hidden batch errors	Surface 3 weeks later in an alert	Zero — validated with the photo in front of you
Batch-to-plant traceability	Rebuilt by hand when there is a scare	Photo + GS1 linked to the batch, automatic
ERP receiving-module licenses	2-3 active users	1 — the rest validate from a tablet
Carrier demurrage	Recurring invoice	Zero
IFS/BRC audit of the batch	Excel rebuilt the night before	Per-batch dossier, automatic, with photo

Impact estimate

Impact estimate for your plant — to be validated with your numbers.

The block below is an estimate to be validated with the specific data of your plant. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.

• Mid-sized plant making food products, with 8-12 daily raw material deliveries (high-gluten flour, seasoning, seeds, packaging film, oil) and batch-to-plant traceability demanded by retail customers and IFS/BRC audits.
• Estimated payback 3-8 months, depending on daily delivery volume and the carrier demurrage documented over recent years.
• Direct savings on ERP licenses: typically a 50-70% reduction in receiving/warehouse module licenses depending on how many people log in today just to register incoming goods, plus removal of the ERP's bulk-capture module cost where it was contracted separately.
• Receiving lead's time recovered ≥ 1 hour/day, going back to their real job (raw material quality control, supplier non-conformance management, coordination with production).
• Carrier wait cut to minutes: direct savings on recurring demurrage invoices and a better relationship with carriers who stop losing their morning at your dock.
• Batch-to-plant traceability complete by photo (audits without archaeological reconstruction). The batch file for the IFS/BRC auditor produces itself.
• Estimate to be validated with your numbers.

And the logistics manager's reasonable doubt

“What if the AI misreads the batch and puts a wrong expiry date into the ERP?” — hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI merely recontextualizes a piece of data from one system to another (reading a GS1-128 label or a delivery note line and translating it into an ERP entry), the best models brought error below 1.5% ^[1]. And even so, what is critical is never decided alone: Connect proposes, the receiving lead validates with one tap, and only then does the data enter the ERP. The photo of the original delivery note is kept as evidence — if someone ever disputes a batch, the original paper is there, with its signature.

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