Today’s exception alert is deliberately blunt and completely honest. Ask MoveTheWheels for tracking alerts and it returns the shipments whose estimated delivery has already passed while they are still in flight — not delivered, not returned, not failed — each with a delay type and a translated one-line message, ordered by how long ago the window closed. It is a real, deterministic query over live freight, and it tells you precisely one thing: that time has run out, after it already ran out.
What is missing is prediction and explanation. The designed version reads the shipment’s tracking timeline and the history of the lane it is running, and produces an expected arrival and a risk score before the delivery window closes — so an exception stops being a post-mortem and becomes a forecast a leader can still act on.
On top of that sits the sweep. Rather than waiting for someone to open the control tower, an agent walks the in-flight book on a schedule, ranks what is genuinely at risk, drafts a resolution for each — re-sequence, re-assign, re-book with a marketplace carrier, warn the receiver — and files it for a human to approve. The queue then arrives ordered by consequence instead of by lateness, which is the difference between a control tower and a list.
The substrate for all of it is already in place, which is why the shape is this specific. Shipments carry origin, destination, estimated and actual delivery timestamps and a tracking event timeline. Delivery jobs carry status history. Routes carry waypoints, distance and fuel cost. The dashboard already computes on-time delivery rate from delivered shipments. What is absent is the model layer that scores risk and the ledger that records what an autonomous run decided and why.
Roadmap and target-experience scenario for a pre-launch platform. The rule-based delay alerts are live; the predicted arrivals, the risk scoring and the autonomous sweep are direction, not current capability.
Ready to make this your story?
