Definition: AI track and trace is software that works the load, not just locates it. It contacts the driver and dispatcher at each milestone, reads what comes back, writes times, status and documents into your TMS, and hands a person only the loads that need judgment.
Step one: the load arrives
The moment a carrier is assigned in your TMS, Chain reads the load: stops, appointments, references, special instructions and contacts. It also reads any conversation that already exists, including messages your reps sent by hand.
Step two: tracking starts from whatever the carrier has
The driver app, an ELD connection, a Samsara or Motive link pasted into the chat, EDI, a carrier API, or a third-party tracking link. Several can run on one load. If nothing starts before pickup, Chain texts the driver a link and keeps a timeline of every send.
Step three: eight checkpoints
| Checkpoint | When | What it does |
|---|---|---|
| Dispatch | Carrier assigned | Driver's number, empty and rolling, ETA to pickup |
| Pre-pickup | Morning of pickup | Welcome, status and tracking in one message |
| At pickup | Near the appointment | Confirms check-in, then loaded status |
| Pickup number | Inside your threshold, tracking live | Releases the confidential number. Only then. |
| In transit | On your cadence | Light check-ins and stale-ETA refreshes |
| Pre-delivery | Hours before | Confirms the driver is on track |
| At delivery | At the appointment | Arrival, unloaded status, every page of the POD |
| Post-delivery | After the appointment | Out-time, the POD once more, escalation if silent |
At each checkpoint the agent reads the load, the tracking and the conversation, then decides whether anything needs to happen at all. It won't ask about arrival while the truck is 200 miles out, and it won't repeat a question a rep already asked.
Step four: it reads what comes back
A text reply becomes an ETA on the right stop. A photo becomes a seal number, a pallet count or a reefer temperature. A POD is checked against the load before it's filed. If it's for a different load, the driver is asked for the right one.
Step five: it writes back
Arrival and departure times, status, next-stop ETA, call-ins in your format, truck and trailer numbers, and the documents themselves, categorized where your billing desk expects them. Those timestamps are often what fires your customer's EDI feed.
Step six: it knows when to stop
When the agent hits something a person should decide, three things happen. An email goes to the reps on the load. Autopilot stops messaging the driver on that checkpoint until a person replies. And the escalation is logged in the thread. Typical triggers: no reply after two messages, an appointment change, a health or personal issue, a question it can't answer, or an ETA falling behind the window.
What it deliberately doesn't do
- Discuss money: rates, detention and lumpers go to a rep.
- Reroute the driver. It asks whether they can reroute themselves.
- Release a pickup number early, or invent one.
- Pretend to be a person if asked.
What to measure
Ask any vendor how they define an automated load and what's in the denominator. Chain publishes three levels of strictness, and across customers at real volume, base no-touch automation runs from 70% to 94%. How we count.