Departments · Logistics and warehouse AI in the Warehouse: Goods In, Delivery Notes and Freight Documents

AI in the warehouse: six tasks at a glance

These are typical workflows, not client projects. In each one the AI prepares and reports differences, and a warehouse operative or the warehouse manager checks and decides. The bridge to the order is built by procurement, the bridge to the invoice by accounting.

1. Read advance shipping notices and freight documents

An advance shipping notice announces what is coming. It arrives as an EDI message, PDF or email, often in a layout of its own for each supplier.

• Input: advance shipping notices, consignment notes and accompanying documents by EDI, email or portal.
• AI prepares: reads supplier, order reference, items, quantities, packages and expected arrival date and matches the notice to the order. If something is missing or does not fit together, it flags it.
• Warehouse checks: the flags, and plans the dock and staff.
• Output: an arrivals list for the day, every line with its order reference.

2. Match goods in with the purchase order

When unloading, what is actually on the pallet counts, not just what is on paper.

• Input: the delivery note, the quantities and barcodes recorded and the purchase order from the ERP.
• AI prepares: compares item, quantity, batch and best-before date with the order, recognises part deliveries and surplus quantities and lists every difference.
• Warehouse checks: the differences on site, and decides whether to accept, partly accept or reject.
• Output: a goods-in draft in the warehouse system or ERP, ready for confirmation by the warehouse.

3. Answer delivery enquiries

"When will my goods arrive?" is the most frequent question to a warehouse, and every answer costs a phone call.

• Input: the enquiry from customers or from sales and the current data from ERP and shipment tracking.
• AI prepares: writes a reply with status, date and the source of the information and marks what it does not know.
• Warehouse checks: replies for exceptions and delays. Simple status replies can go out directly after a trial phase.
• Output: a draft reply or a sent status reply with a log.

4. Document transport damage

When damage is noticed only later, photos, time and the note are missing.

• Input: photos, the note of the person receiving the goods and the delivery note with any reservation.
• AI prepares: assigns photos and notes to the delivery, writes a description of the damage, shows the deadlines for notifying damage and drafts the notice to the carrier.
• Warehouse checks: description and notice, and sends it. Liability and compensation are decided by claims handling.
• Output: a damage file with all documents by date and a draft notice.

5. Record returns

Returns come back with and without paperwork, and each one has to be matched to an order.

• Input: return slip, photo, customer message and the original order.
• AI prepares: matches the return to the order, records reason and condition and suggests the posting (back to stock, inspection, scrap).
• Warehouse checks: the condition of the goods, and confirms the posting.
• Output: a return entry in the system and, where your rules provide for it, a draft credit note for accounting.

6. Reorder suggestions from stock and consumption

Reorders often come too late or too generously because nobody has all the figures in view.

• Input: stock, consumption over recent months, minimum stock and lead times from the ERP.
• AI prepares: code calculates coverage and reorder point from the data. The AI writes a suggestion for each item from them, with the figures it relies on.
• Warehouse and procurement check: the suggestion. Ordering is done only by procurement.
• Output: a list of suggestions with reasons.

Programs and interfaces: connected to what you already use

In the warehouse, data from scanners, warehouse management, ERP and supplier mail comes together. The table lists the usual cases. Which route your system offers, we clarify before the quote.

The names are for orientation and do not mean a partnership. Where a system offers no interface, we work with export and import, a watched folder or email.

Connect, do not replace

Your warehouse does not need a new WMS or a new ERP, and we do not sell either. The AI works next to your warehouse management: it reads what arrives and prepares postings.

• Your warehouse management stays in charge. Stock and postings change only there and only by your people.
• Your routines at the dock stay. Scanning, counting and signing happen as before. The AI works on what would otherwise be typed afterwards.
• One workflow at a time. We start with one step, for example the advance shipping notices or the goods-in matching, and run it alongside the existing work first.
• The way back stays open. All data lives in your systems. Without the AI your warehouse carries on as before.

Approvals and rules

In the warehouse minutes decide dates, and a mistake in quantity costs money later. So before the start we agree what the AI may do.

• Quantity differences: You set tolerances. Within the tolerance the AI only reports, outside it puts the case forward for clarification.
• Postings: Goods in is created as a draft and confirmed by a warehouse operative. The AI never posts by itself.
• Signature and damage note: The signature on the delivery note and the reservation for visible damage stay with the person who accepts the goods. The AI can remind, but cannot sign for them.
• Information to customers: Status replies come from live data and name their source. For exceptions and damage the AI hands over to a person.
• Staff: We do not build systems that monitor individual employees, rate their performance or assign them tasks based on their behaviour. They are high-risk under Annex III of the AI Act, see risk categories.
• Log: every suggestion is recorded with source, time and decision.

The table lists the main sources for orientation. Which rules apply depends on mode of transport, contract and country.

In Germany section 87 BetrVG and in Austria section 96a ArbVG also apply: technical systems that can record the behaviour or performance of employees need the works council's involvement. For personal data in delivery documents the GDPR applies: a data processing agreement (art. 28) and processing in the EU. This overview is general information, not legal advice.

Data: items, storage locations and batches

The quality of the matching depends on the master data the AI works with.

• Item records: Item numbers, EAN, units and packaging units decide whether delivery note and order fit together. If conversions are missing, we report them before we match.
• Storage locations and batches: Batch and best-before details have to be in the system so that the AI can check them.
• Barcodes: Where goods arrive with a shipping label, we read the code instead of typing text.
• Data protection: Names of drivers and contacts in delivery documents are personal data. We process them on servers in the EU, show the AI only what it needs, and agree retention and deletion with you. More under AI and data protection.

Your own apps for the warehouse

Some things are missing not in the warehouse management but on the handheld or the smartphone. We build them as separate programs when standard products do not fit:

• Goods-in app for scanners and smartphones: photograph the delivery note, scan the lines, see differences at once.
• Warehouse app: query stock and location, record transfers and stocktaking, also without a network.
• Delivery note app for drivers: photo and signature at the place of delivery, and the documents land with the right order straight away.

These programs complement your warehouse management, they do not replace it. More under app development.

Where to start

These three questions help with the assessment:

1. How many delivery notes does your goods-in area type up each day, and in what form do they arrive (paper, PDF, EDI)?
2. Which warehouse management and ERP do you use, and do the scanners feed straight into the system?
3. Which step costs the most time: sorting advance notices, matching goods in or answering enquiries?

If you send us the answers, we tell you which workflow is worth doing first and what has to stay with your warehouse. We begin with a trial run alongside the existing work.