What AI means here
Automated track and trace is two jobs. The first is watching: taking positions from wherever they come from, checking every active load against a list of things that can go wrong, and raising the ones that have. That job is rules and clocks. It needs no model, and a vendor who says their AI detects late loads is describing an if-statement in a nicer font.
The second job is acting: deciding what to do about the load that was raised, and doing it. Text the driver. Draft the note to the customer. Propose a new ETA. Read the rate confirmation someone emailed in and make a load out of it. That is where a language model earns its place, because the input is messy and the output is words. It is also where the risk lives, because a model that emails your customer something wrong has spent your credibility, not its own.
So the useful question is not whether a product uses AI. It is which of the two jobs the AI is doing, and who approves it.
The four questions
1. Where does the position come from, and what happens on the loads where it does not? Ask for the list of sources: driver app, text link, ELD integrations by name, carrier share links. Then ask what the board shows for a load with none of them. A product built around one source, usually its own driver app, will report beautiful coverage on the loads that have it and go quiet about the third of your board that does not. Real-time load visibility goes through the four sources and what each one costs.
2. What does it watch for, how often, and can you change the numbers? The list matters more than the map. Stale tracking, a stop that has gone on too long, off the corridor, running late for the appointment, still at the dock past free time, no POD after delivery, a driver who never accepted the dispatch. Ask how often the list is run against every load, and whether the thresholds are yours to set per organisation. A vendor whose thresholds are fixed has decided how your reefer freight works.
3. What does it do about it, and who has to say yes? This is the AI question. For each action the product can take, ask whether it runs on its own, drafts and waits, or only records what it would have done. The right answer is that you choose, per rule. The wrong answer is that everything is automatic, or that nothing is. And ask specifically about messages to your customers: if the product can send one without a person pressing the button, find out how to turn that off before you sign.
4. What does the customer see, and what can they not see? A tracking link for the customer is table stakes now. What separates them is what the page refuses to show: a position older than the stale threshold, a driver's name or phone, an ETA computed from a truck that has gone dark. A page that shows a confident dot for a truck nobody has heard from in three hours is worse than no page.
Red flags
Tracking you opt loads into. If a dispatcher has to decide which loads get watched, the ones that go wrong will be the ones nobody opted in. Every active load, or it is not automated.
No history on the load. Every alert, message and automated action should sit on the load with a time and who was told. If the demo cannot show you one load end to end, the product cannot show it to a carrier during a dispute either.
An ETA with no basis. Ask what the number is built from. If the answer does not include a road route planned for a truck and the driver's remaining hours, it is distance over speed with a decimal point.
A chat window as the main interface. Dispatchers do not want to ask where a load is. They want a board that already knows and only interrupts them for the loads that need a decision.
Coverage claims without a denominator. Ninety-five percent tracked of which loads: the ones with the app, or the board?
A scorecard for the demo
Take this into the call. Score each line yourself; do not let the vendor fill it in.
| Ask | Good answer | Weak answer |
|---|---|---|
| Sources | App, text link, named ELDs, carrier share links | Our app, plus integrations on the roadmap |
| Cadence | Every load checked on a fixed schedule, minutes apart | Real time, unspecified |
| What it watches | A named list, thresholds you set per organisation | Anomalies, detected by AI |
| Autonomy | Per rule: log, draft and wait, or act | Fully automated, or a toggle for everything |
| Customer messages | AI drafts need approval; scheduled status emails are opt-in | No distinction between a draft and an automatic send |
| Audit | Every action on the load with time and recipient | Notifications history |
| Customer page | Hides stale positions and driver identity | Shows the live dot |
| Getting loads in | PDF rate con, spreadsheet, API | Manual entry, integration coming |
How LoadSignal answers
You are reading this on our site, so here is where we stand on each line, including the ones that are not finished.
Sources. Four paths: the driver app with background location, a browser link for one-off drivers, direct Samsara and Motive ELD connections in beta, and supported carrier share links. The team helps set up ELD connections during onboarding.
Cadence and what it watches. Eleven detectors check active loads on a five-minute schedule and on relevant load events. Your organisation configures the supported thresholds, acknowledgment windows and escalation settings.
Autonomy. The Track & Trace Operator is off by default. Each rule can record, draft, or act. Driver reminders can run on their own, subject to your per-rule attempt limit. Customer drafts from the Operator and Pilot AI always wait for approval. Scheduled customer status emails are a separate, opt-in feature.
Audit. Every position, alert, message and action is on the load's timeline with the time and who was told.
Customer page. Shows stops, windows and an estimated arrival. Position is shown only while tracking is active, with its freshness visible. The page does not expose the driver's name or phone number.
Getting loads in. Rate confirmation PDFs, spreadsheet import with column mapping, and a public API with webhooks. Rose Rocket Classic provides one-way load sync. Emailed tracking-link intake can attach a supported carrier link to the matching load.
Access and setup. LoadSignal is in early access for brokerages moving roughly a hundred to a couple of thousand loads a month. Pricing and setup are discussed on a call. The load tracking software guide covers how the wider category divides, and early access is here.