AI Chatbot for Logistics: A Trucking Case and 75% Fewer Calls

An AI chatbot for logistics pays for itself in office silence, not in dashboards. Twenty trucks, two dispatchers and a phone that never stopped ringing after eight in the morning — that was the starting point here.

Seven out of ten incoming calls asked the same thing: where is the truck right now. The dispatcher opened the tracker, read the coordinates, called back. Sometimes mixed up the routes. The customer got nervous, the dispatcher burned out, and nobody was looking for new loads in the meantime.

Familiar chaos in freight? The company shut it down in three days by putting one AI agent on incoming messages in WhatsApp and Telegram. No new hires.

What the AI chatbot for logistics delivered in two months

Dispatcher workload dropped by 75%, the company saved 200,000 roubles a month on support salaries, and reply time went from 10–15 minutes down to 15 seconds. New orders grew 15%, simply because dispatchers finally had time to chase them.

The problem: a dispatcher used as a GPS tracker

The segment is long-haul trucking. Turnover was growing, small things were eating the margin. While a dispatcher sat on the phone, he missed a tender or forgot to send documents to a driver. Pennies on their own, a visible number by the end of the month.

"We were paying professionals to read data off a screen out loud. That is absurd," says the company owner.

The first attempt was to give customers tracker access. It did not work: you have to install the app and remember the password, while writing a message takes seconds. Then came a button-based bot, and customers refused to walk through four menus to learn about one warehouse in Kazan.

The fix: an agent with access to the monitoring system

The agent was built on the ASCN platform and connected to the GPS monitoring system over API. It understands how people actually write: "where is my truck", "when does the load reach the Kazan warehouse", "send me the driver's number". It answers around the clock, at three on a Saturday night as readily as on Monday afternoon.

Raw coordinates never get forwarded. The agent estimates arrival time with traffic and mandatory driver rest factored in. If the truck is being loaded, it says so. If the driver is unreachable, it asks the customer to wait and pings the logistician itself. Nine out of ten location questions close with no human involved.

How a "where is my load" question gets answered

The logic is short. The customer asks in plain words, the agent pulls coordinates and trip status from monitoring, estimates arrival and replies in the same chat. When the data is missing, or the question is not about tracking at all — a paperwork dispute, a claim — the conversation goes to a logistician together with the message history.

The same frame works in other niches where customers wait on a status. We covered it in the cases on the hotel concierge and on lead qualification in real estate. The data source and the scenario change, the mechanics stay the same.

The numbers after two months

The order growth matters more than the savings. Dispatchers did not get faster; they got their day back. Instead of running an information desk, they work on routes, tenders and customers.

"We used to be afraid of running more than 15 trucks, because the dispatchers simply could not cope. Now we see that with an AI agent we can grow to 50 trucks with the same team," the owner says.

What to keep in mind if you repeat this

The GPS link is not optional. An agent without tracker access is a polite answering machine. The whole value comes from it seeing the trucks.

Do not drag the customer into an app. WhatsApp and Telegram are already open on their phone. Every extra installation step costs you half the audience.

Start with tracking, not with everything. "Where is my load" is the most frequent question and the cheapest thing to automate. Close it, see how much time comes back, and only then look at documents and tenders.

Write down where the agent goes quiet. Claims, money disputes, missed deadlines are a human conversation. Define the handover rule before launch rather than after the first argument.

Frequently asked questions

How long did the rollout take?

Three days. Most of that went into the API connection to GPS monitoring and into teaching the agent how customers phrase things. The reply scenario itself comes together faster.

What if the tracker reports with a delay?

The agent states when the data was last updated instead of passing stale coordinates off as current. An honest "the truck was on the M-7 twenty minutes ago" serves the customer better than a neat but wrong arrival time.

Do you need a developer?

Not to build the agent; it is configured in a no-code interface. Code may be needed if your tracker has a closed or unusual API.

What if a customer gets rude or demands compensation?

The agent does not argue and hands the conversation to a logistician. That is an explicit rule, not something the model guesses.

Does this work for a five-truck company?

Yes, though the savings look different. On a small fleet the agent pays off not by cutting headcount but by freeing the single dispatcher from the phone so they can work on loading.

The bottom line

Manual processes do not scale: every new truck adds calls rather than profit. Here the agent took over the routine, and a ceiling of 15 trucks turned into a plan for 50 with the same team.

Ready-made agents for support and tracking are in the ASCN marketplace — you set up the knowledge base and the monitoring connection yourself. If you need help with a tracker API, message a manager on Telegram.