What freight forwarders should look for in AI rate management software

A buying guide written for forwarders, not vendors: the seven things that actually matter in AI rate management — and the marketing claims to discount.

Author
Adam Yaron· Co-founder, Freightools
Updated
Updated
Reading time
6 min read

The thesis

A buying guide written for forwarders, not vendors: the seven things that actually matter in AI rate management — and the marketing claims to discount.

Why this guide exists

Most rate management buying guides are written by vendors. This one is written for forwarders looking at five or six options and trying to figure out which capabilities actually matter — and which ones look impressive in a demo but fail on the floor.

We are one of those vendors (Freightools.ai). The list below is the same list we would use to evaluate ourselves. If a feature is on this list and we do not do it well, we have flagged it explicitly.

The seven things that actually matter

1. Document ingestion that handles your real tariffs

Not the clean carrier sample in the demo. Your real tariffs — the scanned ones, the messy NVOCC rate cards, the supplier emails with the rates buried in a thread.

Test it on the call. Drag in one of your worst PDFs. If the vendor blocks you from doing this, that is your answer.

2. Source-clause traceability

Every extracted field links back to the source document and the clause it came from. Without this, your operations team cannot review at speed, and your finance team cannot defend the price.

Ask: "If a customer disputes a surcharge I quoted, can I open the original tariff page that defined it?"

3. Validity-aware rate retrieval

The system has to return the current valid rate for the lane and equipment, with effective GRIs applied — not a stale or future rate.

Ask: "Show me a lane where two contract versions overlap. What does the system return when I quote during the overlap?"

4. Customer-specific rule encoding

Tier rules, customer rules, lane-specific overrides, time-bound rules. The whole stack of pricing logic forwarders actually use.

Ask: "Walk me through encoding 'Customer A, +6% on Shanghai → Rotterdam, expires Q3'. Then expire it and show me what happens."

5. Margin floor enforcement

The system refuses to quote below floor without a manager override, regardless of who is logged in.

Ask: "Try to override the floor as a junior rep account. What does the system do?"

6. Audit trail

Every change, every quote, every override is logged with who, when, and why. Six months later you can reopen any quote and reconstruct it.

Ask: "Show me a quote from 90 days ago. Tell me which tariff version, which rule, and which surcharge produced each line."

7. A surface the customer can actually use

Self-service quoting that is branded as your business, gated by your rules, and integrated with your operations workflow when bookings are accepted.

Ask: "Can I show this portal to a customer tomorrow without the word 'demo' anywhere on it?"

Claims to discount

These show up in vendor decks. Be skeptical:

  • "99% accuracy on tariff extraction." Accuracy depends on document quality, format, and review threshold. The honest claim is "operations review takes under 10 minutes per clean tariff", not a percentage.
  • "Eliminates your need for ops review." No serious product does this. Ops review is a feature, not a bug.
  • "Replaces your TMS." It almost certainly does not. Most rate management software is meant to feed a TMS, not be one.
  • "AI predicts the best rate for your customer." Predicting the rate is not the hard part — applying your rules to the right tariff is.
  • "Plug-and-play." No software that touches pricing is plug-and-play in a forwarder business. Plan for an honest two-week loading and review cycle for your top lanes.

What to ask in the procurement contract

  • Data residency. Where do tariffs live? Are they shared across customers? (They should not be.)
  • Vendor lock-in. Can you export your structured tariff library at any time, in a usable format?
  • Audit logs. How long are they retained? Can you export them?
  • SLA for ingestion errors. When the system misreads a tariff, who fixes it and how fast?
  • Pricing model. Per-tariff? Per-quote? Per-user? Hidden seats? Make sure you understand the unit.

How to actually run the evaluation

A demo controlled by the vendor will always look good. The way to see the truth is to drive the evaluation yourself, on your data. Before any call, prepare three things: five of your messiest real tariffs, two of your most complicated customer pricing rules, and one old quote you need to be able to reconstruct. Then make the vendor run all three live.

The seven "Ask" prompts earlier in this guide are deliberately phrased as things you do, not things you watch. A capability that survives your documents, your rules, and your audit question is real. A capability that only works on the vendor's curated sample is theater. The difference is invisible in a polished walkthrough and obvious the moment you hand over a scanned NVOCC rate card.

Total cost of ownership beyond the license fee

The sticker price is rarely the real cost. Before you compare quotes, account for the parts that do not appear on the order form: the loading and review effort for your top lanes, the time to encode your customer rules, the integration work to feed your TMS, and the ongoing review queue as tariffs change. A cheaper tool that needs three analysts to keep it fed is not cheaper.

Equally, weigh the exit cost. If you cannot export your structured tariff library in a usable format, the price of leaving is rebuilding everything — which is a cost the vendor is counting on. A fair tool makes your data portable; an extractive one locks it in. Put export rights in the contract, not the sales conversation.

The questions that separate real products from demos

By the time you are in a sales process, every vendor will claim the seven capabilities above. The way to cut through it is to ask questions whose answers cannot be faked in a polished walkthrough.

Ask "show me a value you got wrong, and how the system handled it." A mature product has an uncertainty-handling story — it flags low-confidence fields for review. A weak one pretends it is always right, which means its errors ship silently into your quotes.

Ask "reconstruct a quote from six months ago." If the system cannot show which rate, which validity window, and which rule produced a historical number, you do not have an auditable system — you have a faster way to make undocumented decisions.

Ask "what happens when a carrier changes their template?" The answer reveals whether extraction is genuinely model-driven or a brittle set of hard-coded templates that breaks the first time a carrier reformats a PDF.

Ask "can we export everything?" A vendor confident in their product will hand you your structured data on request. A vendor whose moat is lock-in will get vague. Portability is the single best predictor of how you will be treated as a customer after the contract is signed.

And ask "who reviews the AI's output, and where does that happen in the workflow?" If the answer is "you just trust it", walk away. The right answer puts a human in the loop at the point where a mistake is cheap to catch — before the rate goes live — not after it has propagated into a quarter of quotes.

None of these are gotcha questions. They are the operational realities you will live with every day once the demo glow wears off. A vendor who answers them straight is one you can build on.

Where Freightools.ai stands on this list

We do (1) document ingestion, (2) source traceability, (3) validity-aware retrieval, (4) customer rule encoding, (5) floor enforcement, (6) audit trail, and (7) the customer-facing surface (Instant Quote, branded as you).

What we do not do: replace your TMS; replace your sales team; predict rates we did not load. We are deliberate about that.

If you are running a buying process, book a demo and bring the list above. If we cannot answer all seven on the call, we will tell you.

Summarize this article with AI

Open your favorite assistant and paste the prompt below. Or copy it.

Summarize the article "What freight forwarders should look for in AI rate management software" in 5 plain bullet points for a freight forwarding leader, then list 3 questions I should ask when evaluating freight rate management software. Keep it neutral.

Powered by Freightools.

Frequently asked

Questions freight forwarders ask about AI-native operations

  • How long is a realistic implementation?

    Top-five lanes loaded and reviewed: roughly two weeks. Full coverage of an active forwarder rate library: weeks to months, depending on tariff volume and complexity.

  • Do we need a TMS to use this?

    No. Quotes can be exported, emailed, or pushed to a TMS via integration. Many customers run rate management alongside an existing TMS without replacing it.

  • What happens when a carrier changes their tariff format?

    The ingestion model handles new templates by extracting what it can confidently and flagging the rest for review. You are not blocked waiting for a vendor update.

  • How do we avoid getting locked in to a vendor?

    Make data portability a contract term, not a promise. You should be able to export your full structured tariff library and audit logs in a usable format at any time. If a vendor cannot commit to that in writing, treat the exit cost as part of the price.

See it on your data

Send us one supplier tariff before the call.

We'll show how Tari would structure it and how Miles would quote from it.