VectorLogix
VectorLogix's dispatch team was re-routing drivers manually every time traffic, weather, or a last-minute order changed the picture, and static morning route plans were falling apart by mid-morning. Aivora AI built an agentic route optimization system that continuously monitors conditions and re-plans affected routes automatically, only pulling in a human dispatcher when a decision falls outside its confidence bounds. Fleet-wide, on-time delivery rates rose and dispatcher workload dropped noticeably.
+11 pts
On-time delivery rate
-63%
Manual dispatcher re-routes
-17%
Average driver idle time
6 weeks
Time to first depot pilot
Wherethingsstood
VectorLogix operates a last-mile delivery fleet of around 140 vehicles across a metro region, delivering time-sensitive parcels for regional retail and grocery partners. Routes were planned once each morning using a standard routing algorithm, then manually adjusted throughout the day by a team of six dispatchers whenever traffic incidents, weather, or late order additions disrupted the plan.
Manual re-routing was slow and inconsistent: a dispatcher handling one crisis might miss a smaller disruption forming on another route, and drivers often kept following an outdated plan for 20 minutes or more before a dispatcher caught up with them. Late deliveries were increasingly traced back to these gaps rather than to the original route plan itself.
VectorLogix wanted a system that could watch conditions continuously and re-plan routes on its own within defined guardrails, freeing dispatchers to handle exceptions and customer-facing issues rather than constant manual re-routing.
Whatwebuilt
Aivora AI designed an agentic system built around a planning agent that continuously ingests live traffic data, weather feeds, and VectorLogix's own order management system, re-optimizing affected routes whenever a disruption crosses a defined threshold. The agent operates within guardrails set by VectorLogix's operations team (maximum route deviation, driver hour limits, delivery window commitments) and escalates to a human dispatcher whenever a decision falls outside those bounds.
The system integrates with VectorLogix's existing driver mobile app, pushing updated route segments directly to drivers without requiring a dispatcher to relay the change manually. A six-week prototype phase used simulated traffic disruptions to validate the agent's decision-making before any live routes were touched, followed by a phased rollout starting with 20 vehicles on a single depot.
Full fleet rollout took four months, run in parallel with the existing dispatch process the entire time so dispatchers could compare and build trust in the agent's decisions before fully handing off routine re-routing.
Howtheengagementran
Discovery
Mapped existing dispatch workflows and defined operational guardrails for autonomous re-routing decisions.
Prototype
Built the planning agent and validated its decisions against six weeks of simulated traffic disruptions.
Pilot
Rolled out to 20 vehicles at a single depot, running alongside the existing manual process for comparison.
Rollout
Expanded to the full 140-vehicle fleet over four months, with dispatcher escalation paths refined throughout.
“Our dispatchers used to spend half their day just reacting to traffic. Now the system handles the routine re-routes and hands us only the calls that actually need a person.”
Dana Okafor, Director of Fleet Operations, VectorLogix