Does an AI logistics agent share our freight data with other shippers?
Does an AI logistics agent share your freight data with other shippers?
A distributor of industrial fasteners is close to handing its outbound freight to an AI logistics agent. The logistics manager likes the idea of better rates through a shared carrier network. Then the sales director asks a pointed question in the review meeting: one of the company's direct competitors ships the same lanes out of the same metro area. If both companies use the same agent and the same carriers, can the competitor learn who the distributor ships to, how much it pays or how often it ships? The IT lead adds a second question: does the vendor use the company's orders and emails to train its software?
A properly designed AI logistics agent does not share one shipper's freight data with any other shipper. Each company's orders, rates, customers, documents and messages should stay inside that company's own organization, encrypted, and answerable only to that company's team and the carriers and customers on its own loads. The one thing a network should combine across shippers is freight volume, so that carriers price a lane on a larger, steadier flow. Volume counts together; records do not.
That is the standard to hold any vendor to. It is also something a buyer can test, not just take on trust. The rest of this article explains what data an agent handles, where the separation should sit, what a network does and does not combine, and which questions to put to any vendor before connecting an ERP or forwarding an inbox.
Why the question comes up with agents and not with spreadsheets
A spreadsheet or an in-house system is plainly the company's own. Nobody else has access, so the question never arises. An AI logistics agent is different in two ways. First, it talks to outside parties on the company's behalf: carriers, brokers, consignees and sometimes the company's own customers. Second, a vendor runs the same agent for many shippers at once, often with the same carriers on the same lanes. Both facts are reasonable grounds for concern, and both have clear answers when the system is built correctly.
What "separate" should mean in practice
Separation is not a policy statement. It is a set of concrete behaviors a buyer can check:
- Storage: each shipper's records live in that shipper's own organization and are encrypted.
- Answers: the agent answers a question about a load only when the person asking is a party to that load for that shipper.
- Matching: an inbound email, call or document is matched only to the shipper it belongs to, never to a similar load from another shipper.
- Pooling: the only thing combined across shippers is freight volume for rate negotiation.
- Training: the shipper's data is not used to train the agent.
If a vendor can confirm each of these in writing and show how the agent behaves when a carrier asks an out-of-scope question, the separation is real.
Want to see it with your own team?
Let's walk through how your data is kept separate, with your IT team on the call. You don't need to prepare anything: Cartage maps the workflow with you on the call.
See it on your own loads. Wilson, by Cartage, is an AI logistics agent that quotes, books, tracks, audits and pays for freight with your approved carriers.
Book a demo →What freight data does an AI logistics agent handle?
An AI logistics agent handles the full record of each load: the order, the customer, the rate, the carrier, the documents and every message about it. Knowing what is in that record makes it easier to see what must stay separate and why.
The data on a typical load
Consider a hypothetical outbound LTL shipment from a plant to a customer's warehouse. Before it delivers, the agent will touch:
- The order: order number, customer name, ship-to address, items, weights, pallet count, freight class and requested dates.
- The customer relationship: who the customer is, where it receives, its dock hours and its contacts.
- The rates: every quote requested, every counteroffer, the booked rate and any accessorials.
- The carrier: which approved carrier won the load, its dispatch contacts and its PRO number.
- The documents: bill of lading, rate confirmation, packing list, pallet labels, POD and the carrier invoice.
- The messages: emails, phone calls, texts and chat messages with carriers, customers and the company's own team.
- The payment: the audited invoice amount and when the carrier was paid.
Every one of those items is commercially sensitive. A competitor would find the customer list, the ship-to addresses and the rates useful. A carrier would find another shipper's rates useful in its next negotiation. That is why separation has to cover all of it, not just the rates.
Which data is most sensitive
| Data | Why it matters | Who should see it |
|---|---|---|
| Customer names and ship-to addresses | Reveals the shipper's customer base | The shipper's team; the carrier on that load |
| Booked rates and quotes | Gives leverage in negotiation | The shipper's team; the carrier on its own quote |
| Order volumes and frequency | Shows how much a shipper sells and where | The shipper's team |
| Product details and weights | Can reveal product mix and pricing | The shipper's team; the carrier as needed to move the load |
| Carrier invoices and payments | Shows spend and payment terms | The shipper's team and finance; the carrier on its own invoice |
| Messages and call notes | Contain all of the above in free text | The shipper's team; each party only on its own thread |
Data the carrier legitimately sees
Separation does not mean hiding everything from carriers. A carrier needs the pickup and delivery addresses, the freight details, the reference numbers and the agreed rate to move the load. It sees those for loads it hauls, and only those. What it should never see is another shipper's load, rate or customer, or a load for the same shipper that went to a different carrier.
Data the company's own customers see
Many shippers let an agent tell consignees about delays or delivery times. A customer should hear about its own orders and nothing else. It should not be able to ask about a different customer's shipment, even from the same shipper, and certainly not about another shipper's freight.
How should data separation work between shippers?
Separation should hold at three points: where the data is stored, who the agent answers and what the network combines. A gap at any one of them defeats the other two.
Point 1: Storage inside each shipper's organization
Each shipper's records should sit inside that shipper's own organization in the system, encrypted. A logistics manager at one company should have no path, through search, reports or a question to the agent, to records in another company's organization. The same applies to the agent itself when it works for a shipper: it draws on that shipper's records, its rules, its approved carriers and its history.
This is also what makes an agent useful. When it books a load, it uses the shipper's own approved carriers, its own rate targets and its own approval limits. None of those should be influenced by another shipper's rules or another shipper's private rates.
Point 2: Answers only to parties on the load
Carriers and customers talk to the agent by email, phone and text. That creates a risk a spreadsheet never had: a carrier could ask, deliberately or by mistake, about a load that is not theirs. Picture a hypothetical carrier dispatcher who hauls for two shippers in the same industrial park. He emails the agent: "What's going out of the park on Thursday, and what are they paying?" A correctly built agent answers only about loads for the shipper whose thread he is on and that he is already booked on. It does not list another shipper's loads, confirm their existence or quote their rates.
The same rule covers customers. A consignee asking "where is my order?" gets an answer about its own order. A consignee asking about "the other order on that truck" does not.
Point 3: Only volume crosses the line
A network exists to give carriers a larger, steadier flow of freight to price. Twenty shippers each sending a few loads a month on a lane look, together, like a lane worth placing trucks on. That combined volume is the only thing that should be shared, and it should be shared as volume, not as a list of loads. The network learns that there is demand on a lane. It does not learn, and does not pass on, whose freight it is, where it goes or what each shipper pays.
For more on how pooled rates and the shipper's own contracts sit side by side, see whether pooled freight volume means giving up your carrier contracts.
What two shippers on the same lane share
Consider a hypothetical lane from a metro area to a regional distribution hub. Two competing distributors both ship it every week. Both use the same AI logistics agent. Both happen to use the same approved LTL carrier on Tuesdays.
| Item | Shipper A | Shipper B | Shared? |
|---|---|---|---|
| Orders and customers | In A's organization | In B's organization | No |
| Booked rates | A's own | B's own | No |
| Approved carriers and rankings | A decides | B decides | No |
| Rate targets and approval limits | A's rules | B's rules | No |
| Messages with the carrier | A's threads | B's threads | No |
| Documents and PODs | A's records | B's records | No |
| Freight volume on the lane | Counts toward the lane | Counts toward the lane | Combined to negotiate rates |
The answer to the sales director's question is in the last column. The competitor gains nothing about the distributor's customers, rates or frequency. The only thing both companies contribute to is the total volume that helps carriers price the lane.
What should buyers ask any AI logistics vendor about data?
Ask direct questions, get the answers in writing and test the behavior yourself. Vague answers about "enterprise-grade security" are not enough. Any vendor running an agent for many shippers should be able to answer each of the questions below plainly.
The questions to ask
- Where does our data live, and is it encrypted? The answer should say that each customer's data sits inside its own organization and is encrypted.
- Can anyone outside our company see our records? The answer should name exactly who on the vendor's side can access data and for what purpose.
- What happens if a carrier asks about another shipper's load? The agent should decline and answer only about the carrier's own loads for that shipper.
- What happens if one of our customers asks about another customer's order? Same answer: only its own orders.
- What, if anything, is combined across customers? For a network, the honest answer is freight volume to negotiate rates, and nothing else.
- Is our data used to train your models? The answer should be no.
- Which systems do you connect to, and what do you read and write? IT should see the ERP data flow, the email access and any write-back before approving.
- Is every message and change kept on our loads? A full change history makes later review possible.
For a broader IT checklist that covers access, connections and approvals, see what IT needs before approving an AI logistics agent.
How to test separation in a demo
Words are easy. Behavior is better evidence. In a demo or early setup, ask the vendor to show:
- An email from a carrier asking about a load it is not booked on, and the agent's reply.
- A consignee asking about an order that is not its own.
- A search by the shipper's own team for a lane, showing only that shipper's loads.
- The change history on one load, showing every message and who sent it.
If the vendor cannot show these, treat the separation as unproven.
Common mistakes buyers make
Mistake 1: Asking only about storage. Encrypted storage matters, but an agent that answers questions from outside parties can leak data through its replies. Ask about answers, not just disks.
Mistake 2: Assuming a network means shared records. Some buyers avoid networks because they picture a shared database of loads. A network that combines only volume shares nothing a competitor could use.
Mistake 3: Skipping the training question. If the vendor's models learn from customer emails, a buyer's rates and customer names could shape answers elsewhere. Ask and get a clear no.
Mistake 4: Leaving IT out until the end. IT should see the data flow before the ERP connection is set up, not after the first loads move. Early review is faster than a late objection.
Mistake 5: Forgetting the team's own channels. Many teams talk to the agent in Slack or Microsoft Teams. Make sure those channels sit inside the company's own workspace and that the agent answers there only about the company's own loads.
A short decision guide
| Vendor answer | What it means | Next step |
|---|---|---|
| "Each customer's data is encrypted inside its own organization" | Separation at storage | Ask how answers to outside parties are limited |
| "Carriers only get answers on their own loads" | Separation at answers | Ask to see an out-of-scope request in a demo |
| "We combine volume to negotiate rates" | Network pooling of volume only | Confirm that no load-level data is shared |
| "We use anonymized data to improve the product" | Your data may be used for training | Ask exactly what, and whether you can refuse |
| "We'd have to check" | No clear policy | Treat as a gap until answered in writing |
Which checklist should IT and logistics agree on before connecting data?
IT and logistics should agree on one checklist before the agent reads a single order. It keeps the review short and gives everyone the same answers.
| Item | Question to answer | Owner |
|---|---|---|
| Data scope | Which orders, lanes and customers will the agent see first? | Logistics |
| Order intake | Will orders arrive by email, an ERP connection or CSV/XLSX files? | Logistics and IT |
| ERP connection | What does the agent read, and does it write anything back? | IT |
| Email access | Which inbox or alias does the agent use, and who else can see it? | IT |
| Team channels | Will the team use Slack, Microsoft Teams, email or text? | Logistics |
| Encryption and storage | Is the data encrypted and kept inside the company's organization? | IT |
| Outside answers | What may carriers and customers ask, and what is refused? | Logistics and IT |
| Network | What is combined across shippers, and is it volume only? | Logistics |
| Training | Is company data used to train the agent? | IT |
| Change history | Is every message and change kept on the load? | Logistics and IT |
| Approvals | Who approves new carriers, rates outside limits and payments? | Logistics and finance |
| Excluded loads | Which loads stay out of scope, such as collect or third-party shipments? | Logistics |
Start with a narrow scope
The company controls what the agent sees. A common starting point is a subset of shipments: one plant, one or two key customers or distributors, or every load except those excluded by rule, such as third-party or collect shipments. A narrow scope means less data to review, a faster IT decision and a clear picture of how the agent behaves before more of the business moves onto it. Scope can grow as the workflow is proven.
Record decisions where both teams can find them
Write the answers down once and keep them with the vendor's written responses. When a new person joins IT or logistics, or when the company adds a plant, the checklist shows what was agreed and why. A related view on who owns what after a project ends is in how to keep control of your freight audit data.
Review the first weeks of loads
After the first loads move, IT and logistics should look at a sample of change histories together. Did carriers ask anything out of scope? How did the agent answer? Were any customers told about orders that were not theirs? A short review in the first few weeks catches policy gaps early, while the scope is still small.
How does Wilson, by Cartage, keep each shipper's data separate?
Wilson, by Cartage, is an AI logistics agent that quotes, books, tracks and pays freight for manufacturers and distributors, and it keeps each customer's data encrypted inside that customer's own organization. Traditional logistics software gives teams information. Wilson executes the work required to move shipments forward. Think of it as a free logistics department: what big shippers build over years, Wilson does from day one, in four parts.
The labor. Wilson does the work from the first quote to the final POD: quotes, bookings, paperwork, tracking and carrier follow-ups, by email, phone calls, texts, WhatsApp, Slack and Microsoft Teams. When it talks to a carrier or a customer, it answers only about that party's own loads for that shipper. A carrier or customer cannot ask Wilson about another customer's orders.
The data. Wilson keeps every order, load, quote, rate, document and contact on record as it works, connected to the company's ERP. Wilson replaces the TMS for freight execution and connects to the company's ERP. Each customer's data is encrypted and lives inside that customer's organization. Customer data is not used to train Wilson. Every message, confirmation and change is kept in the shipment's change history, so IT and logistics can review exactly what was said and to whom.
Orders reach Wilson by email, an ERP connection or CSV/XLSX files. Wilson works with SAP, NetSuite, Microsoft Dynamics 365, Oracle, Infor, Epicor, Sage, Acumatica and other ERP environments, with the exact connection depending on the customer's data flow and scope. Write-back to the ERP depends on scope.
The network. Wilson gets better freight rates on every lane because the shipper's volume counts with every other Wilson customer's. The network includes carriers and brokers. Only freight volume is combined to negotiate rates. Two customers shipping the same lane share nothing else: not customers, not rates, not documents, not messages. The company still approves every carrier before Wilson uses it, and books only carriers it has approved. Carriers keep their direct contracts and credit terms with the shipper, and Cartage takes no margin on the freight.
The payments. Wilson audits every carrier invoice against the rate confirmation and the POD before money moves, then pays carriers from the Wilson Wallet, which the shipper funds. Invoices over the agreed rate or outside the rules wait for a person, and a change in a carrier's remit-to or bank details is flagged to the team rather than acted on.
Control stays with the company. Wilson starts with assisted execution, the company sets approval limits, and routine actions can run automatically within those rules as the workflow is proven. The company also controls scope: a subset of shipments, one or two key customers, or every load except those excluded by rule. To see how an agent behaves on real orders before going further, read how to test an AI logistics agent with your own shipment data.
Pricing
Wilson is free on Wilson's network: the shipper pays the freight rate and nothing on top. When a shipper keeps its own carriers, the fee is 2% to 0.5% of what Wilson pays them, falling with volume. There are no seat fees, no setup cost and no contract. Timing depends on scope: most teams start on a few lanes and expand from there.
FAQs
Does Wilson share our data with other customers, including competitors?
No. Wilson, by Cartage, an AI logistics agent, keeps each customer's data encrypted inside that customer's own organization. Two customers shipping the same lane share nothing, and a carrier or customer cannot ask Wilson about another customer's orders. Only freight volume is combined across the network to negotiate rates.
Does an AI logistics agent share freight data between shippers?
It should not. A properly built agent keeps each shipper's orders, rates, customers, documents and messages separate, and combines at most freight volume to negotiate rates. Ask any vendor to confirm this in writing and to show how the agent answers out-of-scope questions.
Is our data used to train Wilson?
No. Customer data is not used to train Wilson. Ask every vendor the same question, because the answer varies across the market.
What does a freight network actually combine across shippers?
In Wilson's case, only freight volume, used to negotiate better rates with carriers and brokers. No load-level records, customer names, rates or documents are shared between shippers.
Can a carrier find out what another shipper pays on the same lane?
Not through Wilson. A carrier gets answers only about its own loads for the shipper it is working with. Another shipper's rates, loads and customers stay in that shipper's organization.
Can one of our customers see another customer's orders?
No. A customer can ask only about its own orders. Wilson does not confirm, describe or quote on any other order, from the same shipper or from another.
Do we give up our carrier contracts to join the network?
No. Carriers keep their direct contracts and credit terms with the shipper, and Cartage takes no margin on the freight. The company decides lane by lane where to use its own carriers and where network rates help.
What should IT review before connecting our ERP?
IT should review which data the agent reads, whether it writes anything back, which inbox and chat channels it uses, how data is encrypted and stored, and who can answer outside parties. A shared checklist between IT and logistics keeps that review short.
Can we limit what data the agent sees at the start?
Yes. The company controls scope: a subset of shipments, one or two key customers or distributors, or every load except those excluded by rule, such as third-party or collect shipments. Scope grows as the workflow is proven.
How can we check what the agent told a carrier?
Every message, confirmation and change is kept in the shipment's change history. IT and logistics can review a sample of loads together to confirm the agent answered only about the right shipments.
Sources
- Cartage, Wilson product FAQ and product scope, October 2026: https://cartage.ai
Let's walk through how your data is kept separate, with your IT team on the call. You don't need to prepare anything: Cartage maps the workflow with you on the call.
