[{"data":1,"prerenderedAt":23},["ShallowReactive",2],{"content:work-warehouse-order-flow:en":3},{"slug":4,"title":5,"sector":6,"scope":7,"summary":8,"challenge":9,"solution":10,"technologies":11,"outcome":18,"details":19},"warehouse-order-flow","From inbox to shipping label","Wholesale and distribution","Integration · Applied AI · Custom software","Connecting email orders, stock management and shipping for a multi-site distributor.","Orders arrived by email and PDF in a dozen formats. Staff retyped them into the ERP, checked stock in a separate spreadsheet and created shipping labels in a carrier portal. Errors surfaced at the customer.","An order intake service extracts order lines from emails and attachments, validates them against the product catalogue and customer master data, and places them in a review queue. Confirmed orders flow into the ERP via its API; stock reservations and shipping labels are created automatically. Exceptions are routed to a person with all context on one screen.",[12,13,14,15,16,17],"Node.js","PostgreSQL","LLM-based extraction with validation","ERP REST API","Carrier APIs","Queue with retries","Illustrative example. In engagements of this type we measure manual touches per order, error rate and order-to-ship time before and after — and report those numbers to the client, not on our website.",[20,21,22],"Extraction runs against a test set of anonymised historical orders; every model change is evaluated before release.","Each order line keeps a link to its source document for traceability.","Idempotent ERP writes: a retried message never creates a duplicate order.",1794663560424]