When human judgment should beat the algorithm
Automatic dispatch is not autopilot. Reserve lets a dispatcher hold a property for a specific vehicle — and the swarm plans around the decision instead of fighting it.
There is a fear that keeps experienced dispatchers skeptical of automatic dispatch, and it is a reasonable one: if the software decides who services what, what happens when the software is wrong? A dispatcher with twenty seasons of judgment knows things the algorithm does not — a customer relationship, a machine's quirk, a situation on the ground. Any dispatch system that cannot accommodate that judgment is not trustworthy. Automatic dispatch has to come with a manual override, or it is not ready for real operations.
Automatic is not autopilot
The right mental model is not a self-driving fleet that removes the human. It is a very capable assistant that handles the mechanical work — the continuous scoring and reassignment that no human could do by hand at that speed — while leaving the judgment calls to the person who has them. The algorithm is excellent at "which of these forty properties should this machine take next given everyone's position." It knows nothing about the fact that the property manager at one condo complex is furious and needs to be first regardless of what the score says.
The goal is not to remove the dispatcher's judgment. It is to remove the dispatcher's clerical work so their judgment has somewhere to go.
Reserve: hold the decision for the right machine
The override that matters most is the ability to reserve a specific property for a specific vehicle. The dispatcher says "this machine gets that property," and the swarm respects it — no other machine can claim the property, and the rest of the fleet plans around it. By default the reserved property is still scored normally for its machine, so the operator picks it up when it is genuinely their best next stop rather than being yanked across town. When timing truly matters, the dispatcher can go one step further and mark the reserved property as that machine's next stop, jumping it to the front of the queue. Critically, none of this is fighting the system. The reservation becomes a fixed point the engine optimizes around, so the human's call and the machine's optimization work together rather than against each other.
This is what makes the earlier discussion of route familiarity practical. Where a property genuinely needs the operator who knows it, or a customer relationship demands a particular sequence, the dispatcher reserves it. Everywhere else, the swarm runs free. You get automatic dispatch with human judgment layered exactly where judgment adds value.
Reopening without losing evidence
Override also means being able to correct the record. A property marked complete or skipped sometimes needs to be reopened — the work was not actually adequate, or circumstances changed. A well-designed system lets a dispatcher reopen a completed or skipped stop without discarding the evidence already captured. The correction is a new state on top of the history, not an erasure of it, which keeps your documentation trail intact even as you fix a mistake.
Override is what makes automation adoptable
There is a broader lesson here about deploying any automation into a domain with experienced practitioners. The automation earns trust not by being perfect but by being overridable. A dispatcher who knows they can pin a property the instant the algorithm gets it wrong will actually let the algorithm run the other ninety-five percent of the time. Take away the override and the same dispatcher fights the system constantly, because they cannot afford to be locked out of a bad call. The escape hatch is not a concession to distrust — it is the precondition for trust.
Key takeaways
- Automatic dispatch is only trustworthy if it comes with a manual override — automatic is not autopilot.
- The system handles the mechanical reassignment; the human keeps the judgment calls the algorithm cannot see.
- Reserve holds a property for a vehicle and the swarm plans around it — optionally as that vehicle's next stop — so the human's call and the optimization work together.
- Reopening a completed or skipped stop should preserve the evidence, and the overridability is exactly what makes the automation adoptable.