How AI shipment tracking actually works (and why it sends a code to your email)

“Where’s my container?” is the most-asked question in logistics — so it’s the first thing people want an AI assistant to answer. Fair. But there’s a gap between a demo that looks like live tracking and one that actually is. This is the honest version of how it works, and why a good one deliberately adds a step before it tells you anything.

Grounded answers vs. live answers

An AI assistant is great at answering from documents you give it — your tariffs, your SOPs, your service FAQ. That’s grounded knowledge, and it needs no integration. Ask it what documents you need to import machinery into Canada and it answers from your material, instantly.

Live shipment status is different. The current position of a container isn’t in a document — it’s in a system that changes by the hour: your TMS, a carrier’s API, or a visibility platform. To answer truthfully, the assistant has to connect to that system. There’s no way around it, and anyone selling “instant live tracking, just paste a script” is selling a guess.

So we’re upfront: on your site, tracking is a “needs a connection to your system” capability, not a plug-and-play one. Quotes intake, customs questions, rates and after-hours capture all work day one; live tracking works once we wire it in.

The two ways to wire it in

When a client wants real tracking, there are two routes:

1. A live API. We query your tracking or TMS system in real time. Freshest data; needs you to expose and authorize an API, and often the underlying data actually originates from carriers or a visibility aggregator (project44, FourKites) or EDI 214 status messages you already collect.
2. A periodic feed. You push us a simple table — booking/container number, status, ETA, and the customer’s email — every so often (a webhook, an SFTP drop, even a scheduled CSV). We cache it per-tenant. Slightly less fresh, far easier to start, and you control exactly what you share.

Most clients start with the feed. It’s the lightest lift and keeps your core systems untouched.

The part that matters: proving it’s yours

Here’s the step that separates a responsible assistant from a leaky one. Shipment status is private. If anyone could type a container number and get its status, ETA and route, that’s a data leak — competitors, bad actors, anyone.

So before revealing anything, the assistant verifies ownership:

1. You ask for a shipment’s status.
2. It asks for the email on the account.
3. If that email is linked to that shipment, it sends a one-time code to the email.
4. You enter the code.
5. Only then does it share the status.

Crucially, it says the same thing whether or not the email matches (“if that shipment is linked to this email, we’ve sent a code”) — so it’s designed not to reveal who owns which shipment to someone fishing. It’s the boring, correct security you’d want protecting your own freight.

That code in your inbox isn’t friction for its own sake. It’s the assistant proving it won’t hand your shipment data to a stranger — which is exactly the reassurance your customers should want.

The honest bottom line

A logistics AI assistant earns most of its keep on day one: deflecting the repetitive quote, customs and rate questions, and capturing leads 24/7 — all grounded, no integration. Live tracking is worth adding, and we can add it — but it’s a connected capability, with a verification step that protects your customers. Anyone who tells you otherwise is skipping the part that keeps your data safe.