Use cases Transport & logistics
Use case · Illustrative
Booking and dispatch agent
Several dispatch agents fill the last places on the same vehicle. Each booking is valid; the vehicle is overbooked, or a stale hold blocks a real one.
01The workflow
AI agents book deliveries, assign vehicles and allocate dock slots across the order system and the dispatch system.
02The rule that must hold
sum(bookings[slot]) ≤ capacity[slot]
A vehicle, dock or delivery slot is never booked beyond its real capacity.
03How it breaks
- A delivery vehicle has one pallet place left for tomorrow’s route.
- Two dispatch agents, serving two customers, read the capacity view within a few milliseconds of each other.
- Both see one place free and both book it. Each booking passes every check.
- Elsewhere, an agent that crashed mid-booking leaves a hold behind, and a legitimate booking is refused with an ordinary-looking error.
- The route is overbooked in one place and under-used in another. Neither shows up as an incident.
The authority pattern holds here as measured: at 10, 50 and 100 concurrent agents it granted exactly the capacity, never more. Read the report →
04Why monitoring misses it
- Task successEvery agent completed; nothing errored or retried.Healthy
- Distributed tracingOne clean trace per action; every span OK.Healthy
- ReconciliationNo lag, no drift, no duplicate keys.Healthy
- The ruleA vehicle, dock or delivery slot is never booked beyond its real capacity.Broken
05Reproduced in the lab
Each failure mode below is reproduced deterministically in the public Agent Control Lab and can be run by its id.
06What we would engineer
Capacity held by one authority that grants before the booking, instead of each agent checking a shared view.
State & Side Effects →Holds with an owner and a deadline, released in a way that never hands back capacity already used.
Recovery & Idempotency →Overbooking checked as its own signal, separate from dispatch health.
Observability & Evaluation →
07What the diagnostic delivers
- The rules this workflow must keep, written down and classified.
- For each rule, the system that can enforce it, or the gap where none can.
- Priority failure modes reproduced where feasible.
- An ordered hardening plan.
Boundary Illustrative engagement, not a client engagement or a measured production result. The failure modes listed are reproduced in the public lab.