# AI for logistics — receiving discrepancies, dock bookings, stock enquiries > What an agent takes on in a warehouse or a distribution centre: receiving discrepancies, dock bookings, availability enquiries, claim preparation. What it never commits to: a stock, a date. **Version HTML :** https://agence-intelligence-artificielle.eu/en/ai-logistics **Langue :** en-GB --- AI by industry # AI for logistics In a warehouse, the task that costs is the discrepancy. Three pallets announced, two received, a delivery note that says something other than the order — and someone spending the afternoon reconstructing what happened, while the docks wait. The information exists in the WMS, in the carrier’s email, on the receiver’s photo. It is never in the same place. [Request the survey](https://agence-intelligence-artificielle.eu/en/contact) - **5 tasks** documented below - **1 limit** no stock or date committed by the agent - **The WMS stays** the agent reads alongside, replaces nothing - **2 to 6 weeks** before the first system is open The cost ## What this industry pays when nothing moves. Ranges, not promises. They come from tasks we meet in this industry and orders of magnitude we measure in the survey — yours may fall outside, in either direction. **Changing nothing** - Receiving discrepancies dealt with within 24 h : **1 in 3 to 1 in 2** - Time to reconstruct a discrepancy : **30 min to 2 h** - Dock slots booked by phone : **most of them** - Availability enquiries per day, handled by hand : **20 to 50** - Carrier claims documented within the deadline : **1 in 2** - Office time absorbed by discrepancies, per day : **2 to 4 h** **The discrepancy system, opened** - Receiving discrepancies dealt with within 24 h : **4 in 5 or more** - Time to reconstruct a discrepancy : **5 to 10 minutes of review** - Dock slots booked by phone : **the exceptions** - Availability enquiries per day, handled by hand : **2 to 5** - Carrier claims documented within the deadline : **nearly all** - Office time given back, per day : **1 h 30 to 3 h** Five tasks ## What we remove, and what we leave. Each task is described with what the agent takes on and what stays with a person. If the second column is empty, we have not understood the industry. 01 ### Reconcile receipt, order and delivery note **What the agent does** It reads the delivery note, the order and what the receiver entered, and produces the discrepancy : what, how many, where it is written. Photos and emails included. **What stays with a person** What to do about it. Accept, refuse, claim : the decision stays with the receiving manager, with the file ready. 02 ### Book dock slots **What the agent does** It proposes slots to the carrier from the schedule, confirms in writing, and warns operations of announced delays. **What stays with a person** Priority between two lorries on the same day. Operations decide, not the diary. 03 ### Answer availability enquiries **What the agent does** ‘Do you have it in stock, and for when?’ It reads the status in the WMS and answers in writing, with the time of the last update. **What stays with a person** The commitment. It gives the stock as read; it reserves nothing and promises no date. 04 ### Prepare a carrier claim **What the agent does** Delivery reservations, photos, consignment note, exchanges : it assembles the chronological file and flags what is missing before the deadline. **What stays with a person** The claim itself, and its amount. The file is ready in an hour; the decision stays with you. 05 ### Keep site and customer instructions **What the agent does** Hours, docks, palletisation constraints, contacts : they live in emails. It extracts them and keeps one sourced sheet per site. **What stays with a person** Validating a new instruction. A palletisation error is a refused delivery. What we do not do ## No stock, no date, no priority is committed by an agent. The agent says what is written in the WMS at the time it read it. It does not reserve, does not promise, does not arbitrate between two lorries. Those acts depend on what it cannot see — a broken forklift, an absent picker, a priority customer. The diagram shows the line. **The method, in one sentence** Ten systems, one survey, one system opened at a time — the one with the shortest payback. [Read the full method](https://agence-intelligence-artificielle.eu/en/#methode), or go back to the [AI agency](https://agence-intelligence-artificielle.eu/en/) page. The tools ## What we open it with, and why here. The choice of tool comes last, and it is reversible. But in this industry, three constraints keep coming back and steer the choice before we even start. **The WMS, as it is** We do not replace it. The agent reads what it exports and writes into what it accepts. If it has no interface, we start from its daily exports. **A circuit visible in n8n or Make** Operations must be able to stop the agent on stocktake day and correct a rule without calling us. A workflow tool can be read ; code cannot. **Written, time-stamped answers** Every availability given carries the time the stock was read. When the claim comes in, that trace is what counts. The objections ## What we hear in this industry, before we start. ### Our receivers enter data on handhelds, not in emails. Does that work ? Yes, the agent starts from what the WMS records, whatever the device. What it adds is the reconciliation with the delivery note and the order — the work someone was doing in the office afterwards. ### And on stocktake day, or when the WMS is down for maintenance ? We switch it off. A stock read during a stocktake is not reliable, and the agent must not answer with it. That is part of the circuit from the start. ### Will carriers accept a slot confirmed by a machine ? They accept a written slot, confirmed within a minute, with a dock number. What they do not accept is calling three times to get one. ### We already have a customer portal. Why do stock enquiries still arrive by email ? Because the portal needs an exact reference and a password. The agent answers ‘the blue cartons from Tuesday’s order’. That is the distance the portal does not cover. Where this industry concentrates ## The areas where we are called most for this. [**Manchester**AI agency in Manchester](https://agence-intelligence-artificielle.eu/en/ai-agency-manchester)[**London**AI agency in London](https://agence-intelligence-artificielle.eu/en/ai-agency-london)[**Birmingham**AI agency in Birmingham](https://agence-intelligence-artificielle.eu/en/ai-agency-birmingham) [**All locations**see the full map](https://agence-intelligence-artificielle.eu/en/locations) Related industries The same tasks, in neighbouring industries : [AI for transport](https://agence-intelligence-artificielle.eu/en/ai-transport), [AI for manufacturing](https://agence-intelligence-artificielle.eu/en/ai-manufacturing), [AI for e-commerce](https://agence-intelligence-artificielle.eu/en/ai-ecommerce). Method notes ## What we have written on this subject. [4 September 2026AI agent or plain automation: the £20,000 question](https://agence-intelligence-artificielle.eu/en/blog/ai-agent-or-plain-automation-the-20k-question) First call — fifteen minutes ## We will tell you which system to open first. Describe the task that costs you the most. We will tell you what is feasible, how long it takes and what it costs. If the answer is no, you will leave with the two reasons why. [Request the survey](https://agence-intelligence-artificielle.eu/en/contact) [Write to the agency](https://agence-intelligence-artificielle.eu/en/contact) --- ## Autres langues - fr-FR : https://agence-intelligence-artificielle.eu/ia-logistique.md - en-GB : https://agence-intelligence-artificielle.eu/en/ai-logistics.md - de-DE : https://agence-intelligence-artificielle.eu/de/ki-logistik.md - es-ES : https://agence-intelligence-artificielle.eu/es/ia-logistica.md - it-IT : https://agence-intelligence-artificielle.eu/it/ai-logistica.md - pt-PT : https://agence-intelligence-artificielle.eu/pt/ia-logistica.md