Answering 'where's my delivery?' with live data: AI support for logistics

Ask anyone who has worked a logistics customer service desk what the job is, and they will tell you: "where is it?", asked a thousand different ways. Where is my pallet. Has it been collected. Why was it not delivered today. Can I have the POD. Industry shorthand calls it WISMO, "where is my order", and in haulage and pallet networks it can make up most of the inbound contact.

It is also the easiest contact to automate well, provided you do it with live data rather than a chatbot guessing. This guide explains how.

Why generic chatbots fail at tracking

A chatbot trained on your website knows your service levels and depot addresses. It does not know where consignment 90311 is. Asked, it will either apologise and hand over, which saves nothing, or, worse, produce a plausible delivery estimate from nowhere.

In logistics, a wrong answer is worse than no answer. A customer told "tomorrow before noon" who gets nothing will phone twice as angry.

The answer: deterministic lookups

The reliable approach is to take the language model out of the critical path for tracking:

1. A pattern matches your consignment reference formats as soon as a customer types one.
2. The system calls your TMS or tracking API directly.
3. The status, latest scan, ETA and tracking link are returned verbatim, as a status card.

In Threadbox these are regex webhooks. No model rewrites a scan time or rounds an ETA. The AI handles the conversation around the lookup, such as greeting the customer, understanding "it was meant to be today" and deciding to escalate, but the facts come straight from your system.

Beyond tracking: the next most common requests

• Proof of delivery. Fetch the POD image and signature for a delivered consignment.
• Redelivery bookings. Offer available slots and book through your scheduling API.
• Missed or late deliveries. Explain what the scans show and raise it with your team, with the consignment attached.
• Damage and shortage claims. Collect the right details first, consignment, photos and description, so the claim arrives complete.
• Quotes. Capture collection and delivery postcodes, pallet count and weight, and pass a structured enquiry to sales.

Handover that ops teams actually like

Automation fails when the human picking up has to start again. Good handover means the agent escalates on clear rules, such as a missed delivery, a damage report, negative sentiment or a customer asking for a person, and the conversation lands in a shared inbox with the consignment, the scan history and a summary already attached. Your team replies from there, and the reply goes back on whichever channel the customer used.

Meet customers where they are

Trade customers email. Consumers increasingly message on WhatsApp. Drivers and depot staff use Slack or Teams. One agent with one knowledge base can serve all of them, with every conversation in one inbox tagged by channel.

Internal knowledge for depots and drivers

Pallet networks run on rules: cut-off times, service levels, restricted postcodes, surcharges and procedures. Train a second agent on network documents and operating procedures, and depot staff can ask it in Slack instead of phoning head office. Browns Distribution, Threadbox's parent company, uses it this way across its haulage network.

Getting started

1. List your consignment reference formats and the API or database that holds tracking status.
2. Set up a regex webhook for each format and test it against real references.
3. Crawl your website and upload service guides, so policy questions are covered.
4. Write escalation rules for missed deliveries, damage and complaints.
5. Launch on the website first, then add email and WhatsApp.

Regex webhooks are included on every Threadbox plan. Custom actions for PODs and redelivery bookings, and WhatsApp, are included from the Growth plan. Data is hosted in the UK/EU. Read the logistics use case or see pricing.