COMPARISON · 02
Can ChatGPT quote freight? Where it helps, and where a quote desk takes over.
Language models are astonishing at reading messy RFQs. The DIY instinct is right about that. What it underestimates is everything after the reading: the rates, the agents, the margin, the approval and the chase. Here is the honest boundary, and a way to test it on your own enquiries.
ChatGPT can read a freight enquiry and write a good reply. It cannot see your rate cards, check a carrier for a live rate, apply your margin rules, email your agents for the lanes you do not hold, or follow the quote up. A quote desk is the system around the model that does those things, and your team approves every reply before it sends.
ChatGPT and templates vs a purpose-built desk
A general chatbot with a good prompt genuinely can read a WhatsApp RFQ, pull out the shipment details and draft a reply that sounds right. If that were the whole job, this page would end here. The rest of the job is where a quote is won or lost:
| The job | ChatGPT + templates | Purpose-built desk |
|---|---|---|
| Reading a messy RFQ | Genuinely good | Genuinely good |
| A detail missing from the enquiry | Guesses, or leaves a blank | Spots what is missing and drafts the email asking for it |
| Your rates | Only what you paste into the chat | Pulls your rate cards and checks the carrier for a live rate |
| Lanes you do not hold | You email your agents yourself | Emails all your agents for the lane in one go, then chases the ones who do not reply |
| Margin | Prices at whatever figure you give it | Drafts with your margin rules and what that customer paid on that lane last time |
| The customer's history | Knows only what you tell it | Looks up the customer's history before it drafts |
| Sending | Whoever copies it out sends it | Your team approves every reply before it sends |
| Follow-up | Back to you | Follow-ups, chasers and reminders happen on their own, on the same conversation |
| A counter-offer | Start again in a new chat | When the customer pushes back, the counter-offer is already drafted |
| An old quote | No idea it has gone stale | When the quote is old, it expires |
Where the DIY setup actually fails
- The invisible miss. A drafted quote that leaves out a charge line looks complete. Nobody notices until the vendor invoice arrives, and by then the margin has gone.
- The stale rate. The chatbot only knows the rate you pasted. If that rate lapsed this morning, the draft is still confident about it.
- The lanes you do not hold. A chatbot cannot ask your agents for a rate. Someone still sends those emails, waits and chases, which is where most of the time on a quote goes.
- The scaling wall. Copy-and-paste quoting is still one enquiry at a time with a person in the middle. The economics of the desk only change when intake to draft runs without that person copying in and out.
How to test any AI on your own quotes
Whether you are trying ChatGPT, a custom GPT or a product like ours, the same six checks separate a good-looking draft from a quote you would sign:
- Use real enquiries. Take ten from last month, including a messy WhatsApp one and one with the weight missing.
- Give it only what it would have had. Does it ask for the missing detail, or guess?
- Check every charge line. Compare the draft with what the job actually cost you. The miss is usually a surcharge or a destination charge.
- Check the margin. Is the price above your floor for that lane and that customer? Our guide to margin floors shows how to set one per lane.
- Check the rate date. Was the rate still valid on the day of the quote?
- Ask what happens next. Who sends it, who chases it, and what happens when the customer comes back with a lower number?
This is not an argument that general AI is weak. It is the same class of technology that powers a purpose-built desk. The difference is everything wrapped around the model: your rates, your agents, your margin rules, the approval and the follow-up. That wrapper is the product.
Questions forwarders ask
Can ChatGPT quote freight?
It can read an enquiry and draft a professional reply. It cannot see your rate cards, check a carrier for a live rate, apply your margin rules or follow the quote up, so the price in the draft is only as good as what you paste in, and a person still has to check every line.
Can ChatGPT calculate chargeable weight?
Yes, if you give it the dimensions, piece count, gross weight and the divisor for the mode. Check the arithmetic against the rule, because a chatbot can state a wrong figure with full confidence. The chargeable weight guide sets out the formula for air, sea and road. See how to calculate chargeable weight.
Is it safe to paste rate cards into ChatGPT?
Check two things first: the provider's current data-use terms for the plan you are on, and your carrier and agent agreements, which often treat contract rates as confidential. Neither question arises when the rates are read where they already live.
When is a general chatbot enough for quoting?
When volume is low, the lanes are few, one person does the pricing and checks every figure, and follow-up is handled by hand. The gap opens with volume: more enquiries, more lanes, more agents to chase and more quotes to follow up.
Use a general chatbot to polish an email, summarise a thread or explain an Incoterm to a new hire. It is excellent at that. The moment the output becomes a priced commitment with your company’s name on it, you need the system around the model: your rates in, your agents asked, your margin applied, a person approving and the follow-up running. That system is what a quote desk is.
Weighing up the other options? Manual quoting vs an AI quote desk sets out the cost per quote, and the freight quoting software landscape lays out every category of tool, with a disclosure on our own row.
See the difference on a real RFQ.
Bring one from your inbox. Watch it priced on your own rate logic with your margin rules applied, not just phrased nicely.
Book the live demo