By Luis Lopez, AI transportation consultant, CEO of Go Hub.io Holdings Corp and subsidiaries, and host of the Freight Guru Podcast
Quote requests rarely arrive in a clean format. One shipper sends a single line: “Need a rate Miami to Atlanta, 6 pallets, Thursday.” Another forwards a long thread with a spreadsheet attached and the real details buried in the third reply. A third pastes a packing list as an image. Pricing teams spend a surprising share of the day just figuring out what the customer is asking for.
RFQ intake is the step between “an email arrived” and “a complete request is ready to price.” It is one of the clearest places for AI to save time in a freight operation, as long as the workflow is designed with guardrails.
What good intake looks like
A complete quote request usually needs:
- Origin and destination, with enough detail to locate them
- Pickup date or window, and any delivery deadline
- Piece count, packaging type and dimensions
- Weight, total or per piece
- Commodity description, and whether it is hazardous
- Equipment or mode needed
- Accessorials such as liftgate, residential, appointment or inside delivery
- Customer and contact information
The goal of AI-assisted intake is to fill this record from whatever the customer sent, flag what is missing, and pass a clean request to the right desk.
Extracting details from emails and attachments
Modern language models are good at reading unstructured text and pulling out specific fields. They can read the body of an email, the earlier messages in a thread, and attachments such as spreadsheets, PDFs and packing lists. Document-reading tools can handle scanned images as well, though accuracy depends on image quality.
Extraction should produce structured output: a clear value for each field, plus a note of where it came from. If the weight came from an attached packing list rather than the email body, the record should say so. That traceability makes review faster and helps catch errors. For a deeper look at how extraction tools differ, see OCR versus AI document extraction in freight.
Detecting missing fields and asking follow-up questions
The most valuable thing an intake system can do is notice what is not there. If a request has origin, destination and pallet count but no weight, the system should flag it right away instead of letting it sit in a queue until a pricer notices.
AI can also draft the follow-up. A short, specific reply works best: “Thanks for the request. To quote accurately, can you confirm the total weight and whether the delivery location has a dock?” A person can approve and send it, or the workflow can send routine questions automatically once the team trusts it. Either way, the customer gets a faster, clearer response than a generic “please send more details.”
Normalizing locations
Customers describe locations in many ways: a full street address, a city and state, a ZIP code, a warehouse nickname, a port or airport code. Intake tools should turn these into a standard format that pricing systems understand. When a location is ambiguous, such as a city name that exists in several states, the system should ask rather than guess.
Normalization also helps spot accessorials. An address that looks residential, a location inside a port, or a delivery to a construction site can all change the price and the equipment needed.
Classifying mode
A single inbox may receive requests for LTL, truckload, drayage, flatbed, expedited and international freight. Classifying the mode early saves time. Some requests make it obvious; others need inference from weight, dimensions, pallet count and timing. A request for a handful of pallets is usually LTL, a full trailer load points to truckload, and a container number signals drayage.
When the mode is unclear, flag it for a person. Some shipments genuinely fit more than one mode, and choosing between them is a pricing decision, not an intake decision.
Routing to the right desk
Once a request is complete and classified, it should go to the team or person who prices that kind of freight. Routing rules can consider mode, region, customer, equipment and urgency. Requests from key customers or with same-day pickups can be pushed to the front of the queue. For broader inbox management, this guide to AI email triage for freight operations covers related patterns.
The pricing handoff, with guardrails
Some teams want intake to flow straight into automated pricing. That can work for simple, repeatable requests, but it needs guardrails:
- Minimum completeness. No automated quote without key fields such as weight, dimensions and dates.
- Margin floors and ceilings. Prices outside an expected range go to a person.
- Restricted freight. Hazmat, oversized, high-value or unusual commodities always get human review.
- Customer rules. Contract rates and customer-specific terms must override generic pricing.
This article on AI pricing guardrails for freight brokers goes further on keeping automated pricing safe.
Tracking response time
Speed matters in quoting, especially on spot freight. Intake systems make response time measurable: when the request arrived, when it was complete, when it was priced, and when the quote went out. Those timestamps show where requests stall. Often the biggest delay is not pricing at all but waiting for missing information that nobody asked for.
Human review still matters
As an AI transportation consultant, I recommend starting every RFQ intake project with full human review. Let the system extract and draft, and have staff confirm every record for the first stretch. Track how often they correct each field. Once accuracy is consistently strong on certain request types, reduce review for those and keep it for the rest.
Review is especially important for weight, dimensions, commodity and hazmat status, because errors there lead to wrong prices, wrong equipment or compliance problems.
Mistakes to avoid
- Letting the model fill gaps with guesses. A missing weight should stay missing until the customer provides it.
- Hiding the source of each field. Reviewers need to see where a value came from.
- Ignoring attachments. Critical details often live in spreadsheets and packing lists.
- Over-automating replies too early. A wrong automatic follow-up damages trust fast.
- Skipping measurement. Without accuracy and response-time tracking, you cannot tell whether the system helps.
Before any quote goes out, it should still pass a final check. This list of things to check on a freight quote is a useful reference for that last step.
For more conversations on freight and AI, subscribe to the Freight Guru Podcast.
About the author: Luis Lopez is a Miami-based AI transportation consultant and logistics entrepreneur, the CEO of Go Hub.io Holdings Corp and subsidiaries, and host of the Freight Guru Podcast.
