Logistics & transport
The same lorry brings the delivery note by email, the CMR on paper and the packing list over WhatsApp, in five languages, and someone keys the references in by hand. Inferana reads that paperwork inside your perimeter, on a flat rate.
What each regulation in your sector requires, and which part of that the deployment solves.
Authorities must accept freight transport information in electronic form, with format and access requirements.
Delivery notes and CMRs extracted into structured fields inside your perimeter, with a record of every query.
Accurate declarations, correct tariff classification and retention of supporting documents for the legal period.
Search across your past files to prepare and cross-check declarations, with every proposal reviewed by a person.
Transport is an essential sector: risk management, supply chain security and incident reporting.
Dedicated EU or on-premise deployment, with access control per team and deployment documentation for your risk analysis.
Legal basis and minimisation when handling driver, tachograph and fleet geolocation data.
Prompts and responses are never stored or used to train models, and inference stays where you deploy it.
The same ones you already use, specific to your sector.
The freight moves on its own. The paperwork is moved by people, by hand, against the clock and on thin margins. This is what you hear on the traffic desk every day:
The delivery note reaches me by email, over WhatsApp and on paper, in five languages.
Traffic desk administrationI have two people keying references out of a PDF all morning.
Administration leadTwo people are retiring and their tariff classification judgement goes with them.
Customs leadWe answer the incident after the customer has called three times.
B2B customer serviceThe haulier invoice does not match the delivery note and we find out when paying.
FinanceOne week to answer a tender with four hundred lanes in it.
CommercialOn low double-digit margins I cannot pay per document.
Managing directorWherever you decide: on European infrastructure we manage, or on-premise on your own infrastructure. In both cases prompts and responses are never stored or used to train models.
The rate is flat and tied to the capacity you deploy. A peak of two million delivery notes in one month costs the same as half a million, as long as the capacity holds. To size it we need your monthly volume, the document types and the response time you can live with.
It depends on the document. Open models read poor scans reasonably well, and there are cases where they fail: carbon copies, stamps over amounts, backlit phone photos. We will not quote an accuracy figure before seeing them. We run a sample of your real documents and measure accuracy per document type before anything is signed.
As support, with human review. It is useful to retrieve how something similar was classified in past files and to draft the entry. Responsibility for the declared heading stays with the declarant, and the system keeps a record of what was consulted to reach it.
With an on-premise deployment, their shipment paperwork stays in your infrastructure and the flow never leaves your network. You can show them the deployment diagram, the policy of not storing prompts or responses, and the access log per user and team.
No. The API is OpenAI-compatible: you integrate from what you already run by swapping the key and the endpoint. Your management systems remain the source of truth and receive the fields already extracted.
Yes, with the same key. There the query does go out to the model provider, so paperwork covered by customer confidentiality clauses is handled with the open models inside your perimeter. It is configured per team.
Yes, through the same flow as the scanner. A phone photo reads worse than a native PDF, especially backlit or with a stamp over the amount, and that gets measured before anything is signed using a sample of your real documents. Whatever does not read cleanly is flagged for review before it reaches the TMS.
The models are open and the API is OpenAI-compatible: you can take the weights with you and point the code elsewhere by changing a key. Your documents, your indexes and the extracted fields are exportable.
In the demo we go through your case: which models you need, where they run and what it takes to meet the regulation that applies to you.
We reply within one business day.
