The starting point
Route planning ran on a cumbersome legacy program. Planning states were passed on as screenshots, appointments agreed by phone. Individual customers were missing from the plan even though they were recorded in the inventory system. On top of that, a booked stock figure did not automatically mean the goods were ready for the next delivery: showroom pieces, items already scheduled and goods still to arrive had to be told apart.
The solution grew out of daily work
Ulrich Reuther at Reuther Media developed an in-house logistics system together with the owner and the team. The existing inventory system was first connected read-only. The decisive piece was an availability logic that reflects how the business actually works: only the main warehouse counts for deliveries, scheduled goods are reserved, and a physical count can override the calculated stock.
The way people use it was also changed with the team. A two-stage pre-planning step was removed after one day because it was not accepted in daily use. Confirming a delivery had to be a single, understandable click. That feedback changed the solution immediately.
The driver's view on the phone
Who is on the way?
For each tour day the map shows the stops in order and, next to them, every open order: green means deliverable and not yet scheduled, blue is planned for another day, grey is still waiting for goods. Anyone living close to the route is called and added to the tour with one click. A tour day fills up with neighbours instead of trips across the city.
What office, warehouse and drivers can use
The application links stock status, routes and staffing. Customers can choose a delivery window via a link they receive. Drivers see their stops on their phone and can send feedback and photos of receipts. A team bot adds a mobile channel, for example for status messages and voice notes.
AI takes on specific tasks: voice notes are transcribed on a server operated in-house. For photos of labels, image recognition suggests article data that the office confirms. Availability and reservation, by contrast, follow fixed rules. An external AI service is used for image recognition; in-house servers do not mean that every step is processed internally.
In addition, the core of an in-house inventory and invoicing system was built: from quotation through purchase order and goods receipt to invoice. Customer history and documents are brought together there. Development took place with AI support in short, successive steps.
What is already established
The logistics application has been live since September 2026. Direct data access revealed an order that had previously been missing; open supplier orders and deliveries still to be invoiced became visible as a work backlog. The rules address the specific causes of incorrect delivery planning that actually occurred.
First feedback from the business after two weeks in use: in the warehouse, the member of staff responsible estimates a saving of at least one hour a day. In the shop, the time spent on delivery planning has roughly halved. Logistics planning as a whole takes noticeably less effort. These are the assessments of the people involved, not a series of measurements; a systematic evaluation will follow. Avoided journeys have not yet been counted, and further practical tests are pending. Status: 28 September 2026.