Asavinter is a road freight forwarder working in France and across Europe. The company arranges transport on behalf of its customers through a network of vetted subcontractors, across three main services: groupage, full loads and express. Two Guided AI agents run there today: one prices incoming rate requests, the other identifies available carriers for confirmed loads.
At Asavinter, a rate request is now priced in under a minute. Before, the same answer waited for an operator to free up, sometimes until the next morning. In a market where the customer asks several forwarders at once, that delay often decides who gets the load.
The challenge
A slow response is a lost opportunity
Every new freight opportunity starts with an email: what price can Asavinter run this lane for? Those requests go out to several forwarders at the same time. Winning the load becomes a race where a good price is not enough, you also have to reply first.
Writing a reply takes work. For each answer, you have to:
- Read an unstructured email, in formats that differ from one customer to the next
- Work out the constraints and requirements: type of goods, loading metres, weight and so on
- Price it against distance, vehicle type, area and urgency
- Write a clear response
Given the volumes involved, that comes to several hours of work a day. On top of it, the inbox has to be watched constantly, because a request spotted an hour too late is usually lost, to another forwarder or carrier.
Covering the load is the second race
Once the order is confirmed, the job is to identify a subcontractor able to run the transport at the best rate. Several sources can turn one up: the internal history of lanes already run, the carrier directories on the freight exchanges, and the carriers who look at an advert without ever calling back.
What Guided does
Prices the request
The team forwards a rate request to the agent. It reads the email, pulls out the origin, destination, the number and format of pallets and any special requirements, then queries Asavinter's pricing API to get a real price rather than an estimate.
It prices all three services, groupage, full load and express, and returns them with their guaranteed transit times, so the customer gets every option. Everything that moves the price is taken into account by the assistant, for example: tail lift at collection or delivery, delivery by appointment, ADR constraints.
It also knows what Asavinter cannot carry. When a request can't be priced, because information is missing or a constraint isn't supported, the agent escalates the file internally rather than guessing.
The reply is sent as a structured quote email, in Asavinter's own format and signature, with the original request attached.
Median turnaround: 54 seconds from request to priced quote in the customer's inbox.
6,500+ rate requests handled in production since go-live.
Finds the carrier
For confirmed loads, Guided posts the offer to the freight exchanges and gathers candidates from three sources: carriers that have already run the lane for Asavinter, carriers looking at the advert, and carriers actively searching that lane.
Those candidates are then ranked by a proprietary machine-learning model, trained on more than a million carrier approaches and the replies they actually got. The model weighs over 50 signals, among them the lane history for that customer, vehicle type, ADR constraints, how urgent the collection is, and each carrier's past response rate. From that it produces a probability of a positive reply within 24 hours, and only candidates above a threshold are contacted.
In practice, the model is about 15 times better at picking out the carriers likely to reply than untargeted outreach.
The agent then comes back with the available carriers and their prices, and the broker decides who gets the load.
4,600+ carrier offers surfaced since launch.
Rolling out AI without a heavy IT project
Asavinter has extended the setup to all of its brokerage teams. The first agent was live within days of the kick-off. Nothing to install: the operators send their requests straight from their TMS, and the replies arrive in their inbox.
What Asavinter says
"Working with Guided has been genuinely useful. They've freed up a lot of time for my teams, helped us bring AI into the way we actually work, and made us far more responsive to our customers, which matters a lot in our business."
Steve, Asavinter CEO
