Logistics & freight
Freight runs on judgment calls. Link each call to its outcome.
Brokers, carriers, forwarders, and 3PLs decide what to quote, which carrier to tender, how to recover a late load, and whether to pay a claim. Where systems link the request, the operator's choice, and the result, they can capture a decision trail that is difficult to reproduce. A large shipment archive alone does not supply that structure.
01Systems you already run
The raw material is already being recorded.
An assessment starts from the software you run today. It asks what those systems hold, never for the records themselves.
- Transportation management
- McLeod, MercuryGate, Turvo, Oracle Transportation Management
- Warehouse management
- Manhattan Active WM, Blue Yonder WMS, Infor WMS
- Telematics and ELD
- Samsara, Motive, Omnitracs
- Load boards and rate tools
- DAT, Truckstop
- Customs and forwarding
- CargoWise, Descartes, Magaya
- Messaging and claims
- EDI 204, 214, and 210 messages, Shared inboxes, Claims logs
02Dataset opportunities
Where the learnable data sits.
Each opportunity follows a decision from input to outcome. That chain, not the number of records, is what makes data useful for training and evaluating AI.
01
Quotes and bids with win or loss
Spot and contract quotes, the rate logic or rep adjustment behind each, counteroffers, and whether the business was won, lost, or later fell off.
- Quote request
- Rate built
- Rep adjustment
- Negotiation
- Won, lost, or fell off
Why it is hard to reproduce
A rate index alone cannot explain a particular negotiation. Linked quotes, adjustments, counteroffers, and win or loss capture choices that may be difficult to reconstruct without the operating history.
AI use cases
- Pricing and negotiation agent training
- Preference data from won and lost quotes
- Benchmarks for quoting under uncertainty
02
Dispatch, load matching, and routing decisions
Load assignments, carrier tenders and rejections, route choices, and how plans changed when conditions did.
- Load available
- Carrier options
- Dispatcher choice
- Tender accepted or rejected
- On-time result
Why it is hard to reproduce
When dispatchers override a proposed plan, the reason and later delivery result can show how they handled constraints. A final route alone omits those alternatives and judgments.
AI use cases
- Agent environments for dispatch and planning
- Evaluating planning models against expert overrides
- Tool-use traces across TMS, telematics, and messaging
03
Exception handling
Delays, missed appointments, damage, and refused loads, with the sequence of calls, re-plans, and customer updates that resolved each one.
- Exception detected
- Diagnosis
- Recovery options
- Action and customer update
- Outcome and cost
Why it is hard to reproduce
Exceptions are where logistics expertise shows, and they are inherently long-tail. Simulations rarely capture the messy, real constraints that experienced operators work around.
AI use cases
- Edge-case and failure-case evaluation
- Multi-step agent training for operations
- Customer communication grounded in real incidents
04
Freight claims and their resolution
Loss, damage, and shortage claims with documents, photos, liability assessments, negotiation, and the amount paid.
- Claim filed
- Evidence gathered
- Liability assessment
- Negotiation
- Paid, reduced, or denied
Why it is hard to reproduce
A resolved claim can pair evidence, including photos, with a liability assessment and a payment or denial. The documented reasoning and negotiated result provide a dispute-resolution task.
AI use cases
- Dispute-resolution evaluation sets
- Multimodal damage assessment with outcomes
- Reasoning benchmarks on liability and documentation
05
Customs classification and entry outcomes
Commercial invoices and packing lists, the broker's tariff classification, agency questions or holds, and the final entry outcome.
- Shipment documents
- Tariff classification
- Entry filed
- Agency query or hold
- Release or correction
Why it is hard to reproduce
Classification reasoning linked to holds, releases, or post-entry corrections can connect document interpretation to an entry outcome. Reproducing that trail requires the underlying documents, decisions, and agency responses.
AI use cases
- Document understanding with expert labels
- Classification reasoning benchmarks
- Agent tasks for trade compliance
03Volume is not the test
What is usually not valuable on its own.
Lots of data is not the point. A record of an expert deciding, and of what happened next, is.
GPS pings and ELD logs on their own
Location traces describe movement. Dispatch choices, exceptions, and delivery outcomes explain the decisions behind it.
Third-party rate data
Rate tables omit your dispatcher's choices, quote revisions, and final win or loss.
Status messages without context
EDI 214 updates describe events, not decisions.
Bills of lading and invoices as documents
Plentiful document types. The value is in the decisions and disputes they connect to.
04Rights and compliance
The sector rules for the dataset.
Rights reviewed before any outreach.
Last reviewed .
How we handle rights and privacy- 01
Freight claims and third-party records
The Carmack Amendment, 49 U.S.C. 14706, may govern certain interstate motor-carrier and freight-forwarder cargo claims; it does not govern every logistics dispute. Load-board and benchmarking licenses may restrict redistribution. A negotiated claim resolution is not a judicial finding.
- 02
Shipper and carrier contracts
Broker-shipper and broker-carrier agreements often treat rates, volumes, and lanes as confidential, and some restrict use of shipment data beyond performing the service. Review them before including customer-identifiable records.
- 03
Driver personal data and telematics
Telematics and ELD records can identify drivers and reveal location patterns. The GDPR and California Consumer Privacy Act may restrict reuse where their territorial and entity requirements apply; labor agreements may also matter. Review must cover notice, legal basis, individual rights, and transfer requirements. Aggregation or pseudonymization does not necessarily make records anonymous or permit licensing.
- 04
Customs broker confidentiality
19 CFR 111.24 generally bars customs brokers from disclosing client business records or connected information without the client's written authorization, subject to specified exceptions. There is no general anonymization exception. Classification reasoning and other client-derived information need an applicable authorization or exception.
- 05
Counterparty identities
Shipper, consignee, and carrier names can reveal commercial relationships. Tokenization and generalized lanes may reduce identification risk, but do not remove contractual confidentiality duties or establish permission to license the remaining information.
05AI use cases
Training examples and evaluation tasks.
Quotes linked to win or loss give pricing agents negotiation examples. Dispatch choices linked to delivery results provide planning tasks. Exception histories and resolved claims add recovery workflows and dispute-resolution cases with recorded outcomes.
06Questions
What owners in this sector ask first.
- Our TMS holds many loads. Where should we start?
- Look for quotes linked to win or loss, tenders linked to acceptance and delivery, exceptions linked to recovery, and claims linked to resolution. Those joins reveal the operator's decision process.
- What makes a quote history useful?
- The first rate, revisions, counteroffers, substitution choices, and the final win or loss show how a negotiation developed. Product and lane context explain the constraints behind each decision.
- Are dispatch logs useful without driver traces?
- Dispatcher choices, carrier options, tender responses, and delivery results form a planning task. Exception notes and recovery actions add the reasoning behind changes to the original plan.
- What do customs classification records contribute?
- Invoices and packing lists linked to tariff classifications, agency questions, and entry outcomes create document-reasoning tasks. The classification rationale explains the broker's judgment.
Find the AI opportunity in your records.
Start with a free assessment of your systems, the decisions recorded, and the outcomes linked to them. You keep ownership.