Route planning and order tracking for a pickup laundry service.
The business collects and returns around 160 orders a day with six vans. Routes were planned the night before in a spreadsheet, and any same-day order meant phoning a driver and hoping.
Bags occasionally went missing between van and plant, which is expensive and hard to explain to a customer.
Our approach
- Routes that re-plan themselvesNew orders are slotted into the best van as they arrive, respecting time windows and van capacity.
- One stop at a timeDrivers see only the next stop, the note from the customer and two big buttons. Everything else is a swipe away.
- Every bag has a tagQR bag tags are scanned at the door, at the plant and on return, so the system always knows where each bag is.
- Customer updatesAutomatic SMS and WhatsApp updates when the van is ten minutes away.
AI around the solver, not in it
Route optimisation uses a standard open-source solver, and AI isn’t needed there. We used AI tools to draft the solver’s constraint configuration and to write simulation scripts that replayed three months of historic orders, so we could compare plans before drivers ever saw them. The dispatcher went through every constraint with our engineers.
- Simulated 14,000 historic orders before launch
- Constraint configs drafted with AI and checked against the dispatcher’s real rules
- Customer message templates approved by staff before going live
What we’d do differently
Drivers ignored the optimised order for the first fortnight because it broke their own habits around lunch and parking. We added a “driver preference” constraint and a short ride-along with each driver, which we now do before launch rather than after.
“Same-day orders used to mean five phone calls. Now the van just goes there.”
Services on this project: Internal tools, Mobile app, AI features.