AI-Driven Planning Across a 140-Location Depot Network
The Challenge
A 140+ location delivery network required three full days of manual planning per cycle — every week. Constraint changes mid-cycle could not be absorbed, causing SLA breaches and reactive re-routing. Suboptimal routing left fleet under-utilised and generated avoidable margin losses.
What We Built
An AI multi-depot planning platform combines demand forecasting, route optimisation and live constraint integration across all 140+ locations simultaneously. Depot capacity, driver availability and traffic data are ingested in parallel. The system re-plans automatically as constraints change throughout the day — no manual re-work required.
Operational Impact
23% reduction in average delivery lead time. 31% improvement in fleet utilisation. Planning cycle compressed from 3 days to 45 minutes — freeing the operations team for strategic work.
Technology Used
Python
LangChain
FastAPI
Azure
PostgreSQL
