Taking the guesswork out of dispatch.

Aurora’s dispatchers were assigning thousands of shipments a day manually — reading orders, weighing vehicle constraints, and picking routes from memory. It worked, until volume grew faster than they could hire.

The routine 80% was eating the time that should have gone to the genuinely hard 20% — the exceptions where a human really does decide better than a machine.

We started with a two-week discovery on real dispatch data — not to build, but to prove the routine cases were predictable enough to automate safely. They were.

• ◆ A routing engine that proposes optimal routes from live orders and constraints.
• ◆ A confidence threshold: anything uncertain is flagged, with context, to a dispatcher.
• ◆ An override loop, so every human correction improves the next suggestion.
• ◆ Monitoring that shows exactly why any route was chosen.

The engine now handles routine routing end to end and surfaces only the exceptions — each one with the data a dispatcher needs to decide in seconds, not minutes. Nothing is a black box: every suggestion can be traced.

Because a human stays in the loop for the hard calls, the team trusted it from week one — and adoption never became a fight.