How Quote Requests Get Read From Email Automatically

In most forwarders, a quote request today is an email that someone reads, copies into a spreadsheet, checks against a rate file and answers. Reading it automatically changes the first half of that chain and leaves the second half with your team.

The path looks the same whether the request comes from a shipper, an overseas agent or your own sales team:

Before anything is automated, agree the field list with your pricers. It is the same list they check in their heads today, written down once. A typical freight request needs these:

The incoterm matters more than it looks. The ICC's Incoterms 2020 rules set out 11 trade terms that allocate costs, risks and obligations between seller and buyer, and each rule lists its costs in one article (A9/B9). Under EXW the seller makes the goods available at its own premises and the buyer arranges the transport from there, so a forwarder quoting the buyer usually prices the pickup too. Under DAP the goods are delivered at the named place of destination. Reading the term correctly decides which legs you quote.

Quote requests do not arrive in one format. One customer writes three lines, another forwards a long thread with the details in the fourth reply, a third attaches an Excel list of twenty shipments, and a tender arrives as a PDF. Rules and templates break on this variety, because every sender writes differently and nobody fills in your form.

This is the job AI does well: reading language rather than a fixed layout. It reads the body and the attachments, finds the origin, destination, cargo and dates wherever they are written, and puts them into the fields of the quote form. It does not decide anything about price.

The plumbing is standard. Microsoft Graph gives a system access to Microsoft 365 mailboxes, including shared mailboxes, and returns each file attachment with its name, content type and size. The Gmail API does the same for Google Workspace and lists read-only mail extraction among its uses. Your customers see no change.

At AJ Worldwide the shared inbox is connected and AI sorts every email that arrives. On a click, it reads a request, attachments included, into the quote form and lists what is missing. The pricer chooses when to read, so the queue stays under the team's control.

Most requests leave something out. The rule that matters is simple: flag it, never guess it. A guessed weight or a guessed incoterm produces a quote that looks finished and is wrong, which is worse than a quote that is late.

The list of gaps is also a memory. It shows which customer always leaves out the weight and which agent never states the incoterm, so sales can raise it once and later requests arrive complete. Until the facts are in, the request waits in its own status instead of sinking in the inbox.

In practice the system lists the gaps next to the request, and the pricer asks the customer for all of them in one reply instead of discovering them one at a time. These are the gaps that come up most:

Once the fields are read, two lookups happen. The sender's address and company are matched to a customer record, which brings in any contract, past quotes on the same lane and who usually handles that account. Then the lane, mode and equipment are matched against the rate book: contracted ocean, trucking and rail rates first, then spot rates where no contract applies. If no valid rate exists, the system leaves the line empty and tells the pricer; it never quietly uses an expired rate.

This only works if the rates live in one place. If they sit in several spreadsheets, the system can read the request perfectly and still hand the pricer a blank rate column. That is why reading email and moving pricing off Excel are usually one project, not two.

Set up the customer record carefully once. The same company writes from different people, different domains and sometimes through an agent; if those are not tied to one customer, past quotes and contract rates stay invisible to the pricer.

The result is a prefilled request in the queue: the facts, the matched rates and the gaps. The pricer reviews a ready quote instead of starting from a blank sheet.

AI has one job here: reading unstructured requests. Everything that involves judgement or a commitment to a customer stays with a person.

We do not quote accuracy figures for reading, because they depend on how your customers write. What protects you is the design: every field the system read is shown to the pricer next to the request, gaps are flagged rather than filled, and no quote leaves without a person approving it.

Start from real mail, not a demo. Collect a few weeks of real requests, including the messy ones, and agree the field list with the pricers who answer them. Then connect the shared inbox and run the reading on live requests while the team keeps quoting the old way for a short, fixed period. What the pricers correct tells you what to fix before you switch.

During the parallel run, watch three things: the fields pricers correct most often, emails that were sorted wrongly or not read, and requests waiting on missing facts. The first improves the field list, the second the sorting, the third the questions you send customers.

Do not ask customers to use a template or a portal. The point is that they keep emailing the way they do today and the work behind the address changes.

At AJ Worldwide this was one part of a larger move: the pricing team went from three spreadsheets and a shared inbox to one system in under five months. Today 14 pricers work every request there, and every quote gets a follow-up date.