AI Logistics Route Optimisation

The planning team at this logistics operator was experienced. They had been building daily routes for years. They knew the roads, the clients, the driver capabilities, and the delivery windows. The problem was that this knowledge lived entirely in their heads, was applied differently by different planners, and took three to four hours to translate into a daily plan every morning.

On the days when the most experienced planner was absent, the routes were noticeably worse. On days with unusual conditions, the manual process could not adapt quickly enough. The inconsistency in plan quality was showing up in on-time delivery rates and in the fuel cost per delivery, but there was no way to isolate it as the cause until we ran the analysis.

We pulled 18 months of historical delivery data, vehicle records, and customer data into a unified dataset. The patterns in the data were clear once it was organised: certain lane and time-window combinations were systematically underperforming. Certain vehicle assignments were repeatedly suboptimal. The planning heuristics that worked well for a smaller operation were not scaling to the current fleet size.

The ML-powered route optimisation system we built accounts for delivery time windows, vehicle capacity, driver shift patterns, and live traffic conditions. It produces a complete daily route plan in minutes. The planning team reviews and approves it rather than building it from scratch. On difficult days, the system produces a better plan faster than the manual process did on easy ones.

What we delivered: Unified historical delivery dataset covering 18 months. ML route optimisation system integrated into the existing dispatch workflow. Operations dashboard tracking planned versus actual performance by route, driver, and customer.

The outcome: Daily route planning time reduced by 67%. On-time delivery rate improved by 23%. Fuel cost per delivery reduced by 11%. The planning team shifted from execution to exception management and the quality of the plan stopped depending on who was in the office that day.