All case studiesLogistics

Agentic Last-Mile Routing Optimizer for 3PL Operator

Multi-agent system that continuously re-optimises delivery routes using real-time traffic feeds, cutting fuel costs by 23%.

⏱ 11 weeks👥 3 engineers + 1 operations analyst

Results

-23% fuel expenditure (measured over 90 days post-launch)

+17% on-time delivery rate

Real-time re-routing decisions in under 500ms

Dispatcher manual interventions reduced from 70/day to 4/day

Integrated with 6 carrier APIs without any carrier-side changes

The challenge

A third-party logistics operator managing 800 daily last-mile deliveries across 3 cities was using static route plans generated the night before. Real-world conditions — traffic, failed deliveries, ad-hoc pickups — caused significant deviation from plan, leading to wasted fuel, late deliveries, and unhappy carrier partners. Dispatchers were manually re-routing 60–80 deliveries per day.

Our approach

1
Integrated real-time traffic via Google Maps Platform and HERE APIs.
2
Built a LangGraph multi-agent system: a Monitor Agent (detects deviation triggers), a Re-routing Agent (calls OR-Tools optimiser), and a Dispatch Agent (pushes updates to carrier apps via 6 carrier APIs).
3
Implemented a React-based dispatcher dashboard with live map view and one-click override.
4
Ran a 3-week shadow mode before live deployment, validating against historical optimal routes.
5
Deployed on GCP Cloud Run with Redis for shared agent state.

Tech stack:LangGraphFastAPIRedisOpenAIOR-ToolsGCP

Want a similar outcome?

Tell us about your challenge and we'll scope a solution within 48 hours.

Start a conversation