Case study
AI-powered dispatch platform
Automated scheduling and dispatch for a growing logistics business, routed around live traffic and driver availability.
Series B freight brokerage, US — 40-person ops team
- 140/day → 12/day
- manual dispatch calls, over 90 days
- 4 weeks
- to the first production release
- 0
- coordinators lost to the change
At a glance
The problem
Dispatch ran on a whiteboard and a phone. Four coordinators spent their shift calling drivers to confirm availability, re-cutting routes by hand when traffic moved, and calling back to correct what they had just assigned.
What we built
An agent that owns the routine assignment path end to end: it reads the incoming load, checks driver hours and current position, proposes an assignment, and confirms it with the driver over the channel they already use. Anything ambiguous is handed to a coordinator with the context already assembled.
The escalation boundary
We designed the hand-back before the assignment logic. The agent escalates when driver hours are within a margin of the legal limit, when a load has handling requirements it has not seen before, or when its own confidence falls below the threshold we set with the ops lead.
What changed
Manual dispatch calls fell from around 140 a day to 12 over the first ninety days. The coordinators did not shrink from four to one; they moved onto customer work that had been queued for months.
Detail
How this one worked
Did it replace the TMS?
No. It runs against the existing TMS through its API. Replatforming would have taken a year this business did not have, and the TMS was never the bottleneck.
What happens when the agent is unsure?
It hands back to a coordinator with the context already assembled. Anything touching legal hours limits escalates by default.
How long until it paid for itself?
Measured against coordinator hours returned to customer work rather than headcount, inside the first quarter.
Bring us the system that has to work.
Tell us what you are building. We come back with scope, a price and a start date within one week.
Talk about your dispatch problem