The Sales Team Stopped
Processing Emails.
They Started Closing Deals.
A global freight forwarding company was losing thousands of hours a year to a single problem: someone had to read every inbound quote request, figure out what was being asked, track down the missing details, and package it up before pricing could even begin. We eliminated that entire workflow with AI. Completely. The sales team never touched another intake email.
Every hour spent
on intake was an hour
not spent closing.
Freight forwarding is a relationship business. The companies that win do so by responding faster, building trust faster, and converting inquiries before a competitor does. Every minute a sales rep spends reading an email to determine whether it is a genuine quote request is a minute not spent on a client conversation.
The company was receiving hundreds of RFQ emails each day across multiple languages and formats. Some arrived as structured forms. Some were dense paragraphs buried in an email chain. Some were missing half the information needed to actually produce a quote. Every single one required a human to read it, interpret it, decide what to do with it, and act accordingly.
The scale of the waste was staggering when measured honestly. Not just the time to read each email. but the cognitive overhead of context switching, the errors from manual data entry, the delays from back and forth clarification chains, and the deals lost to competitors who responded faster because they were not drowning in the same process.
This was harder to solve
than it sounds.
The instinct when hearing "automate email processing" is to imagine a simple rule based filter. That instinct is wrong. and it is why most attempts at this fail.
Freight RFQs do not arrive in a consistent format. A request from a logistics manager in Tokyo looks nothing like one from a procurement officer in Rotterdam. One arrives as a structured PDF attachment. One is three sentences embedded in a reply chain that started three weeks ago. One references cargo by its common name. One uses the technical classification code. One has every detail. One is missing the most important one.
A rule based system collapses on the first exception. And in freight forwarding, exceptions are not edge cases. They are the normal case.
"The system needed to understand freight the way an experienced ops person does. not just read words, but interpret intent, know what was missing, and know how to ask for it in a way that got a response."
misses a step.
We built a system that handles the entire intake workflow from the moment an email arrives to the moment a fully formatted RFQ package lands in the hands of the pricing team. with zero human involvement at any stage.
Every inbound email is read and classified before anything else happens. Genuine RFQs, general inquiries, existing client communications, and noise are sorted instantly. Only real quote requests proceed. Everything else is handled or flagged appropriately without touching a human queue.
AI ClassificationNamed entity recognition and custom freight domain models extract every quote relevant data point from the email body, attachments, and thread history: origin and destination, cargo type and classification, weight and volume, timeline, Incoterms, and any special handling or hazmat requirements. The model understands freight terminology. both formal and informal. across multiple languages.
Named Entity RecognitionThe system cross references extracted data against what is required to produce a viable quote. Every missing field is identified and prioritized. Weight but no volume? Destination but no Incoterms? The system knows what the pricing team needs, and knows which gaps are critical versus which can be reasonably assumed. Nothing moves forward with holes that would stall the quoting process.
Validation LogicWhen critical information is missing, the system generates and sends a targeted follow up email. written in the same language as the original inquiry, matching the tone and formality of the sender, asking only for the specific details that are needed. No generic reply templates. No unnecessary back and forth. A precise, professional ask that gets answered because it respects the sender's time.
Fully AutomatedOnce all required data is confirmed, the system assembles a complete, consistently formatted RFQ package and routes it directly to the sales and pricing team. organized, structured, and ready to price. The team opens a package. They do not open an email. The difference in cognitive load and response time is significant.
Fully AutomatedThree problems most
automation tools never solve.
The difference between a system that works in a demo and one that works in production is the edge cases. These were the three we had to solve to make this work at real freight forwarding scale.
RFQs arrive as plain text, PDF attachments, Excel files, reply chains, forwarded threads, and everything in between. The system needed to extract structured data from fundamentally unstructured input. reliably, not just when conditions are ideal. We built extraction logic that handles the real distribution of email formats, not the idealized version.
A German freight forwarder with international clients receives RFQs in English, German, Dutch, French, Spanish, Mandarin, and more. The system needed to classify, extract, and respond in the language of the sender. not translate everything into a single language and lose the nuance. Domain specific multilingual models made this possible.
A general NLP model does not know that "LCL" means less than container load, that EXW and DDP represent opposite ends of Incoterms responsibility, or that a request for "hazmat" transport triggers an entirely different set of required fields. We built domain specific freight understanding into the model so it operates with the context of an experienced ops professional, not a generic language processor.
Your team is doing
work that AI should be doing.
If any part of your inbound process involves a person reading something to determine what it is and what to do with it, that is a problem worth solving. The technology to eliminate it exists. Tell us what your intake looks like.
Start a Conversation ← Back to all case studiesInbound processing automation is one of the highest ROI AI applications in operations
The freight forwarding use case is distinctive in its complexity, but the underlying problem. high volume inbound requests arriving in unstructured formats that require classification, extraction, follow up, and routing. exists in insurance, legal services, professional services, manufacturing, and beyond. The architecture is the same. The domain knowledge changes.