The forwarder's inbox is your real rate library
Walk through any forwarder office and ask "where do we keep rates?". The answer, after the polite version, is usually: in our inboxes. Carriers send updated rates by email. Procurement forwards the relevant ones to the desk. Reps reply asking for clarification. Three weeks later the deal exists, but it lives in a thread.
That is the buying side of forwarding — and it is the part most software has historically ignored, because it does not look like data. It looks like email.
What a "buying agreement" is, structurally
Strip away the email wrapper and a buying agreement is a small set of fields:
- Origin / destination (port pairs, regions, or door zones)
- Mode and equipment type
- Validity window (from / to)
- Base rate and currency
- Surcharges and exceptions
- Free time and transit time
- Counterparty (carrier or supplier) and the named contact
Every one of those fields is buried somewhere in the email thread, the attachment, or the contract addendum. Hand-keying them into a system is the work most forwarders never quite get around to.
What AI changes
AI changes two specific things, and only those two:
- Extraction at scale. The same model that reads a tariff PDF can read a supplier rate email and pull the fields above, even when the format is sloppy. It links each field back to the email or attachment it came from.
- Ongoing change tracking. When the supplier sends a "FYI new GRI from June 1" email, the system can detect the update, flag the affected agreements, and queue them for review.
That is it. AI does not negotiate the agreement, build the relationship, or replace the buyer. It eliminates the "I have to find the email and re-key the numbers" step.
What this looks like in practice
Inside Freightools.ai, this is how Tari works for the email channel:
- Forward the carrier email to a dedicated inbox. Or connect a shared mailbox so it is automatic.
- Tari extracts the buying agreement fields and links each value to the original email and any attachment.
- Operations sees the extracted record side-by-side with the email and approves with one click — or corrects.
- Once approved, the agreement is live in Miles for quoting and in your customer portal where allowed.
The result: the rates that used to be "in someone's inbox" are now records the rest of the business can act on, with a clean audit trail.
Why this matters more than tariff PDFs
Tariff PDFs are the famous case, but they are the easier case: at least there is one document. The supplier-email channel is the harder case because:
- The "agreement" is spread across multiple messages.
- Updates arrive as informal replies, not new versions of the document.
- The same supplier may send different rates to different desks within your org.
A system that handles the email case well usually handles the PDF case trivially. The reverse is not true.
A short readiness check
If your team is at any of the following, the email-to-agreement pattern is worth investigating:
- A salesperson regularly replies "hold on, let me find the email" before quoting.
- A supplier sends a GRI and you are not sure which deals it affects.
- Operations re-keys the same numbers from the same supplier into a spreadsheet every cycle.
- Finance discovers margin leak weeks after the booking, traced to a surcharge nobody read.
If any of those hit, this is not a process problem — it is a data problem hiding inside email.
The audit trail you did not know you were missing
The quiet benefit of turning emails into structured agreements is not speed — it is defensibility. When a rate lives in a thread, answering "why did we sell at this number in March?" means a forensic search through inboxes, and the rep who handled it may have left. When the same rate lives as a record linked to its source email, the answer is a single lookup: here is the agreement, here is the email it came from, here is who approved it.
That matters most exactly when it is hardest to reconstruct: a margin dispute, a customer claim, a finance reconciliation months after the booking shipped.
Keeping humans in the loop without slowing down
A common worry is that putting AI between the supplier email and the live rate library either slows things down or removes human control. Done well, it does neither. The model handles the mechanical part — reading the thread, pulling the fields, flagging what changed — and a person makes the call that actually matters: is this agreement correct, and do we want it live?
The approval step is deliberately lightweight. A clean extraction is a one-click confirm. A conflicting or low-confidence one is surfaced for a real look. The human spends their attention on judgment, not on re-typing numbers a model already read correctly.
The change-tracking problem most software ignores
The hardest part of supplier rates is not the first email — it is the fortieth. Rates do not arrive once and stay still. A supplier sends a base tariff in January, a "minor GRI" reply in March, a peak-season surcharge note in May, and a one-line correction in June. By midyear, the "current rate" is the sum of five messages that were never designed to be read together.
This is the part traditional rate software quietly ignores, because it assumes rates arrive as clean versioned documents. They do not. They arrive as conversation. A system that only ingests the first attachment captures a snapshot that is wrong within weeks, and worse, it looks authoritative — the stale number is sitting in a database, so everyone trusts it.
Handling the email channel properly means treating every follow-up as a potential revision. When a new message touches an existing agreement, the system should detect which fields it affects, show the before-and-after, and queue the change for an approver. The agreement gains a history: you can see not just the current rate, but how it got there and who signed off on each change. That history is what makes the number defensible six months later.
There is a commercial edge here too. When a supplier sends a GRI, the forwarder who can instantly see every customer quote that depends on the affected lane can re-price proactively — before the margin erodes. The forwarder relying on inbox memory finds out at invoice time, after the damage is done. Same email, two completely different outcomes, decided entirely by whether the change was tracked.
This is why the email-to-agreement pattern is worth more than it first appears. It is not just digitizing rates; it is turning an uncontrolled stream of informal updates into a versioned, auditable record that the whole desk can rely on without anyone having to remember which thread held the latest number.
See it on your real inbox
If you want to see how the email-to-agreement flow handles your supplier mix, book a demo. We can run it on a sample of your real (anonymized) thread, on the call.